Use of NMR Binding Interaction Mapping Techniques to Examine

These maps showed that BNP and BOH inserted into the surfactant headgroup's major chiral groove and interacted predominately with the leucine chiral c...
0 downloads 0 Views 202KB Size
J. Phys. Chem. B 2006, 110, 17359-17369

17359

Use of NMR Binding Interaction Mapping Techniques to Examine Interactions of Chiral Molecules with Molecular Micelles Kevin F. Morris,*,†,‡ Bridget A. Becker,†,§ Bertha C. Valle,| Isiah M. Warner,| and Cynthia K. Larive†,§ Departments of Chemistry, UniVersity of Kansas, Lawrence, Kansas 66045, Carthage College, Kenosha, Wisconsin 53140, Louisiana State UniVersity, Baton Rouge, Louisiana 70803, and UniVersity of California, RiVerside, California 92521 ReceiVed: May 3, 2006; In Final Form: July 5, 2006

NMR spectroscopy was used to investigate the association of four chiral molecules with the molecular micelle poly(sodium N-undecanoyl-L-leucylvalinate) (poly(SULV)). Adding poly(SULV) to the background electrolyte in electrokinetic chromatography (EKC) allows enantiomeric resolution to be achieved because enantiomers interact differentially with the chiral centers on the micelle headgroups as they both move in the electric field. Pulsed field gradient diffusion experiments were used to measure molecular micelle association constants for enantiomers of each analyte. These association constants were consistent with EKC elution order for the compounds 1,1′-binaphthyl-2,2′-diyl hydrogen phosphate (BNP), 1,1′-bi-2-naphthol (BOH), and Troger’s base. In addition, nuclear Overhauser enhancement spectroscopy, nuclear Overhauser effect difference, and intermolecular cross relaxation diffusion experiments were used to generate binding interaction maps for each chiral analyte. These maps showed that BNP and BOH inserted into the surfactant headgroup’s major chiral groove and interacted predominately with the leucine chiral center. (+)-Troger’s base was also found to insert into the major chiral groove. However, this compound instead interacted with the valine chiral atom. In diffusion experiments with long diffusion times, the linearized diffusion plots for each analyte-molecular micelle mixture showed curvature characteristic of intermolecular cross relaxation. The magnitude of this effect scaled linearly with the analytes’ free energies of binding.

Introduction Group epitope or binding interaction mapping techniques are used to identify the regions of a small molecule that interact with a protein receptor or other macromolecule. This atomic scale information is especially useful if the molecule or ligand in question is a drug candidate.1 In ligand-receptor maps, percentages are assigned to ligand hydrogens to represent the strength of their interaction with receptor atoms. Higher percentages indicate a stronger interaction with the receptor or identify ligand protons that are especially close to atoms of the macromolecule. NMR techniques are used widely in these mapping applications because resonances from each nonequivalent ligand proton can be observed and because the nuclear Overhauser effect (NOE) allows through-space interactions between the ligand and macromolecule to be detected. The goal of this research was to establish whether NMR-based epitope mapping techniques, developed in large part to characterize proteinligand interactions, can be used to investigate the binding of chiral analytes to dipeptide-terminated molecular micelles (MMs). In these applications, we report the results of the NMR analyses as binding interaction maps and reserve the designation epitope maps for protein-ligand systems. Chiral MMs are added to the background electrolyte in electrokinetic chromatography (EKC) to achieve enantiomeric * To whom correspondence should be addressed. E-mail: kmorris@ carthage.edu. Phone: (262) 552-5481. Fax: (262) 551-6208. † University of Kansas. ‡ Carthage College. § University of California. | Louisiana State University.

resolution of racemic mixtures. Enantiomers interact differentially with the chiral centers on the headgroup of the surfactant as they both move through a capillary under the influence of an electric field.2 As a result, the two enantiomers may migrate with different average velocities to achieve chiral resolution. In molecular micelles, surfactants are covalently linked to one another within the hydrophobic core. These covalent attachments prevent dynamic exchange of monomers between the micelle and free solution, providing the MM with the rigidity and stability needed to facilitate chiral resolution in EKC under a variety of elution conditions.2 Studies have been conducted to characterize the physical properties of dipeptide-terminated MMs and to probe their interactions with chiral compounds.2-15 In general, MMs containing a dipeptide headgroup with the larger of the two amino acids in the N-terminal position have been shown to provide better chiral resolution than polymers containing a single amino acid.3,4 NMR studies have shown that, in solution, the headgroups of the molecular micelles adopt a folded structure containing a chiral pocket into which the analyte enantiomers insert.8 The present study builds upon previous NMR work involving MMs.8,9,16 Nuclear Overhauser enhancement spectroscopy (NOESY), NOE difference, and pulsed field gradient NMR experiments were used to generate binding interaction maps for four chiral analytes binding to the molecular micelle poly(sodium N-undecanoyl-L-leucylvalinate) (poly(SULV)). These maps allow us to probe the mechanism of chiral recognition in EKC and establish the best NMR-based mapping technique for this system. While this study focuses on chiral analyte-MM interactions, the NMR methods employed may also be used to

10.1021/jp0627224 CCC: $33.50 © 2006 American Chemical Society Published on Web 08/17/2006

17360 J. Phys. Chem. B, Vol. 110, No. 35, 2006 examine the interactions between an analyte and other EKC mobile phase modifiers such as micelles, dendrimers, or cyclodextrins.16 In NOESY and NOE difference experiments, the strength of the NOE interaction is proportional to the distance between the protons experiencing the NOE, provided that the correlation times for all of the analyte’s protons are comparable in the MMbound state. Therefore, by comparing the relative NOE strengths between all of the analyte protons and a given atom of the MM, the analyte protons that are closest to and are thus likely to be experiencing the strongest interactions with the molecular micelle can be identified. In NOESY, the strength of the NOE interaction is determined from cross-peak volumes. NOE difference experiments provide analogous information from the areas of the resonances in a difference spectrum generated by subtracting spectra with on and off resonance irradiation of selected MM protons. Mapping of intermolecular interactions with NOESY has been used by Mayer et al. to identify the ligands in a library of oligosaccharides that bind to a protein receptor.17 The NOE difference mapping experiments are similar to the saturation transfer difference technique used to characterize protein-ligand interactions.18-24 Pulsed field gradient (PFG) NMR diffusion studies were also carried out because these experiments can be used to both measure analyte-MM association constants and generate binding interaction maps.25-27 When an analyte binds to a macromolecule such as a protein or MM, the small molecule’s diffusion coefficient changes dramatically because in the bound state the analyte diffuses at a rate characteristic of the larger macromolecule. The analyte’s association constant can be calculated by measuring the change in its diffusion coefficient in the presence of the micelle or protein.27 Group binding interaction mapping is also possible with PFG-NMR when experiments utilizing a version of the stimulated echo pulse sequence are employed. In the stimulated echo experiment, gradient pulses are used to spatially encode the magnetization that is then stored longitudinally during the diffusion period.28 If long (>500 ms) storage periods are employed, intermolecular cross relaxation can occur between macromolecule and analyte protons within 5 Å. This effect is strongest for protons that are in close spatial proximity. Analyte resonances experiencing intermolecular cross relaxation during the longitudinal storage period in a diffusion experiment exhibit deviations from a purely exponential decay of their signal intensity with increasing gradient pulse area. The magnitudes of these deviations are quantified and used to generate group epitope maps.25,26 The diffusion-based mapping method typically requires shorter experiment times than two-dimensional experiments such as NOESY and has the advantage that information about an analyte’s binding affinity and its binding interaction map can be obtained in a single experiment. To date, however, the intermolecular cross relaxation diffusion experiment has only been applied to the study of ligand-protein interactions.25-27 The study reported here was undertaken, in part, to determine if the technique could be extended to investigations of analyte binding to molecular micelles for the compounds shown in Figure 1, while providing information about binding interactions important for electrophoretic separations incorporating MMs. In this system, the cross relaxation effects are expected to be smaller and more challenging to detect because the association constants of the analyte-MM complexes are typically much smaller than those of the protein-ligand systems previously investigated using this approach.

Morris et al. Experimental Details Materials and Methods. The R and S enantiomers of propranolol, 1,1′-binaphthyl-2,2′-diyl hydrogen phosphate (BNP), and 1,1′-bi-2-naphthol (BOH) as well as the (+) isomer of Troger’s base (TB), the protein R1 acid glycoprotein (AGP), (cyclohexylamino)propanesulfonate, and DCl were purchased from Sigma-Aldrich. Sodium phosphate dibasic and sodium borate were from, respectively, Mallinckrodt and Baker, Inc. and Fisher Scientific. Wako Pure Chemical Industries, Ltd. provided the (R)-warfarin. Deuterium oxide (99.9 atom % D) was provided by Cambridge Isotope Labs. All compounds were used as received. The molecular micelle poly(SULV) was synthesized and purified by use of a method previously reported by Wang and Warner.2 The molecular structures and proton designations for poly(SULV) and the analytes (i.e., BOH, BNP, propranolol, TB, and warfarin) are shown in Figure 1. Chiral analyte-MM solutions for BOH, BNP, and propranolol containing 5.0 mM analyte and 50.0 mM equivalent monomer concentration of the polymer were prepared gravimetrically. The Troger’s base-MM mixture contained respective analyte and polymer equivalent monomer concentrations of 5.0 and 25.0 mM. These concentrations were chosen to mimic the conditions under which the EKC separations were performed. Relatively low analyte concentrations were also necessary to ensure the 1:1 analyte to MM binding stoichiometry assumed in the analysis of the diffusion experiments described below. The BNP and BOH solutions were prepared in a 50.0 mM sodium borate buffer at pD 10.2, and the propranolol solutions were prepared in a pD 8.5 buffer containing 300.0 mM (cyclohexylamino)propanesulfonate and 50.0 mM sodium borate. A 30.0 mM phosphate buffer at pD 7.0 was used in the Troger’s base experiments. These conditions were similar to those used in previous EKC analyses employing these analytes. The mixture containing the protein AGP and the analyte (R)warfarin was prepared in D2O by diluting concentrated stock solutions. The pD of the protein-ligand solution was adjusted to 7.17 with DCl. The concentrations of the AGP and (R)warfarin were 0.142 and 1.62 mM, respectively. NMR Experiments. All NMR experiments were performed using a Varian INOVA 600 MHz spectrometer and a 5 mm inverse triple-resonance (1H, 13C, 15N) probe equipped with actively shielded triple-axis gradients. The probe’s gradient coil constant was calibrated at 49.8 G/cm at 100% gradient strength using a solution of 10.0 mM β-cyclodextrin in D2O.29 The temperature was regulated at 298 K in all experiments. In the NOE difference experiments, selective saturation of MM resonances was achieved by applying a series of 50 ms Gaussian-shaped rf pulses, each separated by a 1.00 ms delay for a total saturation period of 2.00 s. Selective saturation was followed by a hard 90° rf pulse and then a spin lock pulse of 25 ms. The spin lock is incorporated into the sequence as a T1F filter in STD experiments to remove the protein signal in protein-ligand applications. Here identical difference spectra were recorded with and without application of the spin lock pulse. A proton spectrum containing 32180 points with a spectral width of 8503.2 Hz was collected. In the next scan, the saturation frequency was moved 5000 Hz and an off-resonance saturation spectrum was collected with the same number of points and spectral width. Difference spectra were obtained by internally subtracting on and off resonance spectra after each scan to minimize subtraction artifacts from magnetic field inhomogeniety or temperature changes. A total of 512 transients (following 16 steady-state transients) were collected for each saturation frequency, and 3 Hz line broadening was used when the

Interactions of Chiral Molecules with MMs

J. Phys. Chem. B, Vol. 110, No. 35, 2006 17361

Figure 1. Molecular structures of (a) poly(sodium N-undecanoyl-L-leucylvalinate) (poly(SULV)), (b) 1,1′-bi-2-naphthol (BOH), (c) 1,1′-binaphthyl2,2′-diyl hydrogen phosphate (BNP), (d) propranolol, (e) Troger’s base, and (f) warfarin.

difference spectra were processed. The method described by Mayer and Meyer was used to generate binding interaction maps from the NOE difference spectra.18-20 Each analyte resonance in the difference spectrum was integrated, a value of 100% was assigned to the largest integral, and the other integrals were then expressed as a percentage of the maximum value. In principle the NOE difference maps can be generated by either irradiating the MM resonances and monitoring the analyte difference peaks or irradiating the analyte peaks and comparing the resulting MM resonances in the difference spectrum. However, reproducible binding maps could only be generated when the MM protons were irradiated. Consequently, only the NOE difference maps generated by irradiating the MM resonances will be presented. Phase-sensitive NOESY spectra were acquired by coaddition of 24-32 transients (4 steady-state transients) measured into 1024 F2 data points for each of the 256 increments in F1. Linear prediction was used to extend the data set by 200 points in F1. The data were zero filled to 1K × 2K, a π/2-shifted sine-bellsquared apodization function was applied in F1 and F2, and the data set was Fourier transformed in both dimensions. The NOESY mixing time was 400 ms. Binding interaction maps for each of the well-resolved analyte resonances were generated from the NOESY spectra using projections taken parallel to the F1 axis along each of the data points defining the resonance in the F2 dimension. These projections were coadded to yield one-

dimensional spectra containing the analyte peak of interest and peaks from protons that showed an NOE cross-peak to the analyte resonance. In these projected spectra, each peak’s area is proportional to the strength of its NOE with the analyte resonance along which the projection was taken. The largest integral was assigned a value of 100%, and the other integrals were expressed as a percentage of this maximum value. Group binding maps generated with these methods are shown in Tables 2 and 3. The bipolar pulse pair stimulated echo (BPPSTE) pulse sequence was used in all pulsed field gradient NMR diffusion measurements.28 In this experiment, the intensity of a resonance, I, decays exponentially with increasing magnetic field gradient strength, G, according to eq 1.

[

(

I ) I0 exp -(γGδ)2 ∆ -

δ τ - D 3 2

)]

(1)

I0 is the resonance intensity with zero gradient, γ is the magnetogyric ratio, δ is the duration of the gradient pulses, ∆ is the diffusion time, τ is the short delay between the bipolar gradients, and D is the diffusion coefficient. For the analyte-MM mixtures, δ and τ were 4.0 ms and 70 µs, respectively. In the AGP-(R)-warfarin experiment, δ was 3.00 ms and τ was 70 µs. In PFG-NMR, intermolecular cross relaxation effects are only observed at longer diffusion times, so ∆ values of 600-800

17362 J. Phys. Chem. B, Vol. 110, No. 35, 2006

Morris et al.

TABLE 1: PFG-NMR Diffusion Results for Enantiomers of BOH, BNP, TB, and Propranolol (Prop)

analyte (R)-BOH (S)-BOH (R)-BNP (S)-BNP (+)-TB (-)-TB (R)-Prop (S)-Prop

Dfree × 1010 (m2 s-1)

Dobs × 1010 (m2 s-1)

Db (cm2 s-1) (m2 s-1)

fb

K (M-1)

5.40 ( 0.05

1.12 ( 0.01 1.27 ( 0.01 1.77 ( 0.05 1.64 ( 0.01 1.51 ( 0.02 1.58 ( 0.02 1.12 ( 0.01 1.06 ( 0.01

1.00 ( 0.01 1.06 ( 0.01 1.11 ( 0.03 1.08 ( 0.01 1.16 ( 0.02 1.13 ( 0.01 1.03 ( 0.01 1.02 ( 0.01

0.955 ( 0.020 0.952 ( 0.020 0.837 ( 0.019 0.862 ( 0.016 0.931 ( 0.012 0.917 ( 0.012 0.980 ( 0.008 0.991 ( 0.007

425 ( 11 393 ( 11 103 ( 5 125 ( 3 538 ( 28 444 ( 9 993 ( 52 2191 ( 26

5.14 ( 0.05 6.27 ( 0.04 5.54 ( 0.02

ms were employed in these experiments.25-27 Each analyteMM data set consisted of 21-25 spectra containing of 22704 points and with a spectral width of 5997 Hz. The intensities of the gradient pulses were incrementally increased from 8.9 to 42.7 G/cm throughout the data sets. In the AGP-(R)-warfarin experiments, 15 spectra were collected in two different diffusion experiments with solutions containing the protein-ligand mixture and only the protein. The gradient range in each AGP experiment was 4.63-19.0 G/cm. The spectra collected for the sample containing only the protein were subtracted from the protein-ligand spectra at each corresponding gradient. After data acquisition, the free induction decays were apodized with 2.0 Hz line broadening, Fourier transformed, and baseline corrected. Analyte and micelle resonances were then integrated, and plots were prepared of the natural logarithm of the peak area vs (γGδ)2(∆ - δ/3 τ/2). The resulting curves were then fit with the program Kaleidagraph (Synergy Software) to the following second-order polynomial:

I(q) ) A - βq + κq2

(2)

In eq 2, A is a constant, β corresponds to the diffusion coefficient in the absence of curvature, and q represents (γGδ)2(∆ - δ/3 - τ/2). The second-order coefficient κ was used to quantify the curvature or nonlinearity of each plot using the method described by Yan et al.25 Binding interaction maps were generated from resulting curve fits by assigning the proton with the highest κ a value of 100%. The κ values for the other resonances were then expressed as a percentage of the maximum value. In the diffusion experiments used to measure the association constants, the diffusion coefficients of the analytes were measured in free solution and in mixtures containing each analyte enantiomer and the MM. Throughout this part of the study, ∆, δ, and τ were 75.0, 2.00, and 0.20 ms, respectively. For experiments containing only the analyte, each PFG-NMR data set contained 10 spectra with the gradients incremented from 4.63 to 17.1 G/cm, while in the analyses with the analyte-MM mixtures 15 spectra were collected with gradients ranging from 4.62 to 44.5 G/cm. After data acquisition, the free induction decays were processed as described above and the analyte aromatic and MM Hc resonances were integrated. Plots were prepared of the natural logarithm of the peak area versus q, and a linear regression analysis was used to calculate the diffusion coefficients of the analyte and molecular micelle. In the MM-analyte mixtures, the analyte molecules undergo fast exchange on the NMR time scale between the bound and free states. Therefore, the analyte diffusion coefficient measured in the mixtures, Dobs, is the weighted average of the free, Dfree, and micelle-bound values, i.e.

Dobs ) fbDb + (1 - fb)Dfree

(3)

In eq 3, Db is the molecular micelle diffusion coefficient and fb is the mole fraction of analyte molecules bound to the MM. The analyte-MM association constant, K, is given by eq 4,

K)

fb (1 - fb)[SULV]

(4)

where [SULV] is the equivalent monomer concentration for the molecular micelle. Equations 3 and 4 assume a 1:1 stoichiometry for the polymer-analyte complex, and therefore should be used only at low analyte concentrations where 1:1 binding can be assumed. For each chiral analyte, separate PFG-NMR diffusion experiments were performed for the analyte enantiomers, and enantiomeric association constants (KR and KS) were calculated for each molecule. Results and Discussion Association Constants Observed for Analyte-MM Complexes. In Table 1, Dfree, Dobs, Db, fb, and the enantiomeric association constants for BOH, BNP, Troger’s base, and propranolol are presented. In the case of BOH, the association constant for the R enantiomer (KR ) 425 ( 11) was greater than that of the S enantiomer (KS ) 393 ( 11). The larger K observed for the R enantiomer suggests that it has the stronger interaction with the MM. In an EKC study using the same experimental conditions as were used here, the elution order of the BOH enantiomers was (S)-BOH before (R)-BOH.30 Since the MM is anionic, the later eluting enantiomer (e.g., (R)-BOH) experiences a stronger chiral interaction with the MM than the enantiomer with the shorter retention time (e.g., the S enantiomer).11,30 Both PFG-NMR and EKC measurements are in agreement that (R)-BOH interacts more strongly with the MM compared with (S)-BOH. Analogous results were obtained using a third analytical technique, fluorescence anisotropy (FA), using the same set of experimental parameters as were used in the PFG-NMR and the EKC experiments.30 The fluorescence anistropies for the (R)-BOH-MM complex were consistently higher than those of the (S)-BOH-MM complex. The rationale for this result is that the (R)-BOH-MM complex rotates more slowly and for a longer period of time (larger anisotropy) than the (S)-BOH-MM complex (lower anisotropy) due to a higher binding affinity.30 Similar results where three analytical techniques, PFG-NMR, EKC, and FA, were in agreement as to which enantiomer showed a stronger interaction with the MM were observed with the enantiomers of BNP.30 In an EKC analysis of a Troger’s base-poly(SULV) mixture, the analyte’s (+) enantiomer eluted after the (-) enantiomer.30 This result is also consistent with the PFG-NMR association constants which showed that the (+) enantiomer of Troger’s base had the larger MM association constant. The method described above for measuring enantiomeric association constants is most accurate when the fb value is less

Interactions of Chiral Molecules with MMs

J. Phys. Chem. B, Vol. 110, No. 35, 2006 17363

Figure 2. A portion of the NOESY spectrum of a mixture containing (R)-BOH and poly(SULV).

Figure 3. (R)-BOH resonances observed in the NOE difference spectrum of an (R)-BOH-poly(SULV) mixture upon irradiation of the (a) MM valine HR, (b) MM hydrocarbon chain protons (Hc), and (c) MM methyl resonance at 0.90 ppm.

than 0.99. As the fb value increases further, the PFG-NMR method becomes less accurate because the denominator in eq 4 contains the term 1- fb. At large fraction bound values, small errors in fb often produce large changes in the association constant. Therefore, measurements of K values above ca. 2000 M-1 become difficult.

NOESY, NOE difference, and diffusion experiments were conducted and binding maps were generated only for the analyte enantiomer with the higher association constant. Therefore, binding interaction maps will be presented for (R)-BOH, (S)BNP, (S)-propranolol, and (+)-Troger’s base. (R)-BOH-Poly(SULV) Binding Interaction Maps. A portion of the NOESY spectrum of the (R)-BOH-poly(SULV) mixture is shown in Figure 2. Strong negative NOE cross-peaks were observed between the analyte resonances and those of the macromolecule. Negative NOEs are expected for macromolecules such as molecular micelles with relatively long correlation times. While bound, the analyte molecules take on the motional properties of the MM. Therefore, negative NOEs are expected for the analyte protons as well. Figure 3 shows a representative set of NOE difference spectra for a mixture containing (R)-BOH and poly(SULV). Trace (a) displays the (R)-BOH resonances in the difference spectrum obtained by irradiating the polymer valine R-proton. Note that the strongest difference peaks and thus the strongest NOEs are observed for analyte resonances H3 (H3′), H4 (H4′), and H8 (H8′). In the difference spectrum shown in trace (b), the MM Hc resonances were irradiated. The strongest difference peaks are now observed for (R)-BOH protons H7 (H7′) and H5 (H5′)/ H6 (H6′). Finally, trace (c) shows the (R)-BOH region of the difference spectrum generated upon irradiation of the MM dipeptide methyl resonances at 0.90 ppm. In this difference spectrum, the (R)-BOH intensity pattern is identical to that observed in a one-dimensional proton spectrum of the analyte. The generation of group binding interaction maps from these spectra is described above. The binding interaction maps for (R)-BOH and the MM generated with both NOESY and NOE difference experiments are shown in Table 2. There is reasonable agreement between the NOE difference and NOESY maps for the (R)-BOH-poly(SULV) complex. If we rank the percentages from the NOE difference maps, the following results are obtained for the valine, H4 (H4′) > H8 (H8′) > H3 (H3′) > H7 (H7′) > H5 (H5′), H6 (H6′), and the leucine, H4 (H4′) > H8 (H8′) > H3 (H3′) > H7 (H7′) > H5 (H5′), H6 (H6′), irradiations. This order correlates reasonably well with the ranked order we obtain from the NOESY maps. Both techniques agree that protons H4 (H4′) and H5 (H5′)/H6 (H6′), respectively, have the strongest and weakest enhancements. However, differences are seen for protons with intermediate values.

TABLE 2: Binding Interaction Mapping Results for Binaphthyl Compounds analyte resonance (%) experiment

H3 (H3′)

H4 (H4′)

H5 (H5′)

H6 (H6′)

H7 (H7′)

H8 (H8′)

NOESY map to leucine HR NOESY map to valine HR NOE difference map to leucine HR NOE difference map to valine HR

(R)-BOH-Poly(SULV) Maps to Chiral Centers 73 100 41 96 100 28 82 100 48 79 100 41

41 28 48 41

90 90 56 49

90 92 93 83

NOESY map to micelle Hc NOE difference map to micelle Hc

(R)-BOH-Poly(SULV) Maps to Micelle Methylene Chain 50 61 89 62 57 100

89 100

100 85

61 55

NOESY map to leucine HR NOESY map to valine HR NOE difference map to leucine HR NOE difference map to valine HR

(S)-BNP-Poly(SULV) Maps to Chiral Centers 100 81 72 100 75 46 88 100 45 88 100 41

69 43 43 49

72 52 53 51

87 75 86 84

100 100

73 82

49 61

(S)-BNP-Poly(SULV) Maps to Micelle Methylene Chain NOESY map to micelle Hc 15 34 64 NOE difference map to Hc: micelle irradiated 63 59 92

17364 J. Phys. Chem. B, Vol. 110, No. 35, 2006

Figure 4. (a) A comparison of the intensities of the NOESY crosspeaks between the (R)-BOH resonances and the MM leucine and valine HR. (b) A comparison of the (R)-BOH integrals in the NOE difference spectrum generated by irradiating the MM leucine and valine R-protons.

Figure 4 compares the NOE interactions between the (R)BOH protons and the micelle leucine and valine HR resonances. Figure 4a was generated from the cross-peaks in the NOESY spectrum, while Figure 4b was generated from the results of the NOE difference experiment. In the NOESY analysis (Figure 4a), stronger analyte NOEs are observed for the micelle leucine R-proton. This result suggests that the (R)-BOH analyte interacts primarily with the leucine chiral center. Analogous results were obtained with the analyte (S)-BNP (data not shown). However, in the NOE difference experiment, consistently stronger difference peaks for either the leucine or valine HR are not observed. With an NOE difference irradiation time of 2.0 s, irradiating one of the R-protons likely spreads saturation throughout the surfactant headgroup via spin diffusion, thus preventing us from identifying the analyte binding site. Table 2 also compares the binding interaction maps to the MM hydrocarbon chain resonances (Hc) generated with NOE difference and NOESY experiments. In both experiments, the

Morris et al. percentages fall into two distinct categories with those of H7 (H7′), H6 (H6′), and H5 (H5′) consistently larger than those of H4 (H4′), H3 (H3′), and H8 (H8′). This result suggests that, upon binding to the molecular micelle, (R)-BOH protons H7 (H7′), H6 (H6′), and H5 (H5′) are closer to or point toward the hydrocarbon chain, and H3 (H3′), H4 (H4′), and H8 (H8′) point toward the chiral centers on the dipeptide headgroup. Both maps suggest that the analyte binds to the MM by inserting into the major chiral groove formed by the dipeptide headgroup and the hydrocarbon chain.8,30 The analyte protons H3 (H3′), H4 (H4′), and H8 (H8′) which are nearest the -OH group point toward the chiral centers, and the analyte protons on the opposite side of the molecule, namely, H5 (H5′), H6 (H6′), and H7 (H7′), point toward the surfactant’s hydrocarbon chain. (S)-BNP-Poly(SULV) Binding Interaction Maps. The MM binding constant for the (S)-BNP-poly(SULV) complex (K ) 125) was lower than the corresponding value for (R)-BOH (K ) 425). Despite the smaller K value, analyte resonances were detected in the NOE difference experiment when the MM resonances were irradiated and NOESY cross-peaks were observed between the BNP resonances and the MM methylene and dipeptide headgroup R-protons. As with (R)-BOH, the NOESY cross-peaks between the analyte protons and the MM leucine chiral center were consistently larger than the corresponding cross-peaks to the valine HR, suggesting that (S)-BNP also interacts predominatly with the MM leucine chiral center. The (S)-BNP NOESY and NOE difference maps are presented in Table 2. A comparison of the (S)-BNP binding interaction maps to the maps of the MM chiral centers generated with NOE difference and NOESY experiments shows that, in both experiments, the percentages fall into two distinct categories, with H4 (H4′), H3 (H3′), and H8 (H8′) giving higher percentages than H5 (H5′), H6 (H6′), and H7 (H7′). The two methods are in disagreement as to which proton should be assigned 100% in the binding map. In NOESY, the largest enhancement was to H3 (H3′), while, in NOE difference experiments, the largest difference peak corresponded to H4 (H4′). Table 2 also compares the NOE difference and NOESY maps to the maps to the MM Hc resonances. Both the NOESY and NOE difference maps show percentages again falling into two categories with those of H5 (H5′), H6 (H6′), and H7 (H7′) now larger than those of H4 (H4′), H3 (H3′), and H8 (H8′). The maps also agree that protons H6 (H6′) should be set at 100%. In the NOESY-based map, however, we see a larger variation in the percentages compared to those of the NOE difference map. Nonetheless, the (S)-BNP maps generated with these two methods suggest that this analyte’s interactions with poly(SULV) are quite similar to those of (R)-BOH. (S)-BNP also interacts predominantly with the leucine chiral center and inserts into the micelle headgroup’s major chiral groove. Both NOESY and NOE difference experiments show that the analyte atoms closest to the BNP phosphate group show stronger interactions with the MM R-protons, while the protons on the opposite side of the molecule show stronger NOE interactions with the micelle’s hydrocarbon chain. (+)-TB-Poly(SULV) Binding Interaction Maps. In Table 3, the (+)-TB-poly(SULV) maps from NOESY and NOE difference experiments are compared. NOESY cross-peaks were not detected between the (+)-TB protons and the MM leucine R-proton, suggesting that the site of chiral recognition for this analyte is the valine amino acid. The NOESY map can be interpreted in a manner similar to that of the maps generated for the (R)-BOH-poly(SULV) and (S)-BNP-poly(SULV) com-

Interactions of Chiral Molecules with MMs

J. Phys. Chem. B, Vol. 110, No. 35, 2006 17365

TABLE 3: Binding Interaction Mapping Results for the Analytes Troger’s Base and Propranolol analyte resonance (%) experiment

H1 (H7)

H3 (H9)

H4 ( H10)

51

81

Troger’s Base-Poly(SULV) Maps to Micelle Methylene Chain NOESY map to micelle Hc 100 73 NOE difference map to Hc: micelle irradiated 100 70

71 68

Troger’s Base-Poly(SULV) Maps to Chiral Centers 100

NOESY map to valine HR

analyte resonance (%) experiment NOESY map to valine HR NOESY map to micelle Hc

H2 100

H5′

H6′

H7′

H8′

Propranolol-Poly(SULV) Maps to Chiral Centers 89 20 20 26

H2′

H3′

H4′

19

36

26

77

49

46

Propranolol-Poly(SULV) Maps to Micelle Methylene Chain 28 58 52 52 100

plexes; namely, the largest percentages are observed for protons closest to the analyte’s chiral plane and H-bond donor/acceptor atoms. Therefore, (+)-TB also inserts into the polymer headgroup’s major chiral groove with the hydrogen atoms closest to the hydrogen bond donor/acceptor atoms pointing toward the polymer’s valine chiral center. In the NOE difference analysis, saturation of the polymer valine HR resonance produced difference peaks for each of the three (+)-TB aromatic protons. However, the signal-to-noise ratio for the analyte resonances in the difference spectrum was not sufficiently large to generate quantitative binding interaction maps. Finally, when the NOESY and NOE difference maps to the MM hydrocarbon chain are compared, we see excellent agreement between the two methods. (S)-Propranolol-Poly(SULV) Binding Interaction Maps. NOE difference binding interaction maps could not be generated for the analyte (S)-propranolol because the analyte’s side chain proton resonances overlapped with the MM peaks. As a result, the micelle peaks could not be saturated without also affecting the analyte resonances. Therefore, only the NOESY-based maps are presented for (S)-propranolol. The (S)-propranolol binding interaction maps are shown in Table 3. In the NOESY spectrum, the analyte proton at the chiral center (H2) can be resolved. The propranolol-poly(SULV) NOESY-generated maps suggest that the analyte inserts into a MM chiral pocket, placing the analyte protons near its chiral center close to the MM R-protons and the analyte protons on the opposite side of the molecule close to the polymer side chain. No map could be generated for the leucine R-proton due to spectral overlap. Table 3 also shows the NOESY-generated map to the micelle -CH2 protons. The smallest percentages are observed for protons at or near the chiral center and -OH moiety, and the largest percentages are observed for aromatic protons H5′ and H6′ on the opposite side of the molecule. This result suggests that on average protons H5′ and H6′ are closest to the micelle hydrocarbon chain, while the analyte protons near the propranolol hydroxide group are farthest from the micelle Hc protons. Comparison of NOESY and NOE Difference Methods. The results presented in Tables 2 and 3 show that both NOESY and NOE difference experiments can be used to map the binding interactions between chiral analytes and molecular micelles. Both methods lead to an analogous picture of analyte binding and largely agree on the atoms that should be assigned the highest and lowest percentages in the binding interaction maps. General agreement between the two techniques is expected because the NOESY experiment is equivalent to sequentially applying a selective π pulse at every chemical shift in the spectrum.

The differences seen in the maps provided by the two techniques can be attributed to a variety of factors. Small (i.e., 10% or less) differences in the percentages such as those observed in the (R)-BOH map to the leucine HR for all protons except H7 (H7′) are probably not physically significant. Other larger differences may be attributable to differential analyte proton-spin lattice relaxation during the 2.0 s irradiation time in the difference experiments. When long irradiation times are employed and two analyte nuclei have different spin-lattice relaxation times, a stronger NOE will be observed to the nucleus with the longer relaxation time, even if the two nuclei are equally close to the saturated macromolecule atom. The 2.0 s irradiation time in NOE difference experiments also spreads saturation over a number of atoms via spin diffusion, thus preventing detection of differential interactions at or near specific MM atoms. In contrast, spin diffusion does not appear to adversely affect the NOESY experiments with mixing times of 400 ms. Binding interaction mapping with NOESY also provides a global picture of all intermolecular NOEs present in an analytemacromolecule mixture and may be the method of choice in applications where the binding site it not known or when it is not clear which macromolecule resonance should be saturated. In other applications, binding interaction mapping with NOE or saturation transfer difference experiments may be preferable because the difference experiments can be done in a relatively short time, require little postprocessing of the data, and since the difference spectra only contain resonances that experience NOEs or saturation transfer, spectral interpretation is often relatively straightforward. The one-dimensional difference experiments also do not contain t1 noise, which can often interfere with cross-peak and NOE quantization in NOESY. Therefore, we conclude that the NOESY and NOE difference maps generated for the chiral analytes are most useful in identifying the analyte protons experiencing the strongest and weakest interactions with the MM. Further work on systems where the binding map is known will be needed to optimize irradiation and mixing times to determine the experimental parameters that lead to maximum agreement between the two techniques. Binding Interaction Maps from PFG-NMR. To assess the applicability of NMR diffusion experiments to detect analyteMM interactions through intermolecular cross relaxation, we first present results for a ligand-protein mixture containing (R)warfarin and the protein AGP. The ligand is similar in size and structure to the analytes used in the MM study. However, the cross relaxation effects in the NMR diffusion experiment are more readily identifiable. Therefore, a comparison of the analyte-MM PFG-NMR results to those obtained with the warfarin-AGP mixture can be made to assess whether signifi-

17366 J. Phys. Chem. B, Vol. 110, No. 35, 2006

Morris et al.

Figure 5. (a) PFG-NMR diffusion data set and ligand epitope map for the (R)-warfarin-AGP mixture. (b) PFG-NMR diffusion data for the (+)-TB-poly(SULV) mixture recorded using the BPPSTE pulse sequence with ∆ ) 600 ms, δ ) 2.0 ms, and G ranging from 8.9 to 34.6 G/cm. The data sets for each (+)-TB resonance were fit to both linear and second-degree polynomial equations. (c) PFG-NMR residual plots for (left) warfin-AGP and (right) (+)-TB-poly(SULV) PFG-NMR data sets.

cant intermolecular cross relaxation is present in the analyteMM experiments. Plots of the natural logarithm of the peak area vs q (see eq 2) for the (R)-warfarin resonances are shown in Figure 5a. Table 4 lists the diffusion coefficients and κ values for each (R)-warfarin resonance along with the R2 values for the linear and polynomial fits. Warfarin proton H12 exhibits the largest κ value, suggesting that this ligand proton is closest to the protein atoms in the binding site. The warfarin H4 proton has the smallest κ value. Therefore, this proton is least affected by intermolecular cross relaxation. The other ligand protons ranked in order of decreasing κ values are H2 > H1, H3 > H5, H6, H7, H8, H9. These κ values were used to generate the epitope map shown in Table 4. Figure 5b shows a plot of the natural logarithm of the peak area versus q for the poly(SULV)-(+)-Troger’s base mixture with a diffusion time of 600 ms. The data are fit to both a straight line and a second-order polynomial. The curvature observed in this data set (Figure 5b) is modest compared to the curvature seen in the protein-ligand analysis shown in Figure 5a. This result demonstrates that the intermolecular cross relaxation effects observed in the chiral analyte-MM mixtures are smaller than those present in the protein-ligand system. Similar PFG-NMR plots were obtained with (R)-BOH and (S)propranolol. Table 4 also summarizes the apparent diffusion

coefficients, β, and κ values from the PFG-NMR analyses of mixtures containing the MM and (R)-BOH, (+)-TB, or (S)propranolol along with the R2 values for the linear and polynomial fits of the analyte-MM data sets. In the (S)-BNPpoly(SULV) PFG-NMR experiments, no significant curvature was observed for any of the (S)-BNP resonances. This observation is presumably due to the smaller association constant of (S)-BNP compared to the other analytes investigated. Both Figure 5 and Table 4 show that the curvature observed in the protein-ligand analysis was much larger than that observed in the MM-analyte experiments. In fact, the largest κ value calculated for the protein-ligand mixture is an order of magnitude greater than the corresponding values for the MM systems. However, the analyte-MM and the protein-ligand PFG-NMR results share features that suggest the modest curvature observed in the analyte-MM experiments resulted from intermolecular cross relaxation. First, in the PFG-NMR analyses with each chiral analyte, different analyte resonances showed correspondingly different κ values. For example, in the (+)-TB-poly(SULV) analysis, the κ values range from 4.84 × 10-14 to 10.6 × 10-14 cm4 s-2 for resonances H7 (H1) and H10 (H4), respectively. A similar range is observed for the (S)propranolol-poly(SULV) complex, while a slightly narrower κ range (6.13 × 10-14 to 10.7 × 10-14 cm4 s-2) is observed in

Interactions of Chiral Molecules with MMs

J. Phys. Chem. B, Vol. 110, No. 35, 2006 17367

TABLE 4: Summary of Intermolecular Cross Relaxation Diffusion Mapping Results

mixture (R)-warfarin-AGP

proton(s)

H1, H3 H2 H4 H5, H6, H8, H9 H7 H12 (R)-BOH-poly(SULV) H3 (H3′) H4 (H4′) H5(H5′) H6 (H6′) H7 (H7′) H8 (H8′) (+)-TB-poly(SULV) H10 (H4′) H9 (H3) H7 (H1) (S)-propranolol-poly(SULV) H2′ H3′ H4′ H5′ H6′ H7′ H8′ a

β(polynomial fit) Dappa(linear fit) × 106 × 106 κ × 1014 κ R2poly (cm2s-1) (cm4 s-2) (%) (cm2 s-1) R2(polynomial fit) R2(linear fit) R2linear 4.80 4.60 4.84 4.90 4.90 4.42 1.62 1.55 1.53 1.74 1.54 1.66 1.44 1.43 1.51 1.25 1.15 1.15 1.07 1.08 1.05 1.19

127 149 72.3 119 119 162 7.65 10.7 9.91 9.65 7.61 6.13 6.42 4.84 10.6 9.65 4.13 4.13 2.58 4.37 3.58 6.93

78 92 45 73 73 100 71 100 93 90 71 57 61 46 100 100 43 43 27 45 37 72

3.61 3.21 4.16 4.02 3.79 2.90 1.48 1.37 1.37 1.58 1.42 1.56 1.32 1.33 1.29 1.02 1.05 1.05 1.02 0.983 0.961 1.02

0.9995 0.9990 0.9946 0.9989 0.9990 0.9992 0.9994 0.9992 0.9996 0.9984 0.9996 0.9992 0.9996 0.9994 0.9996 0.9988 0.9998 0.9998 0.9998 0.9998 0.9996 0.9992

0.9925 0.9870 0.9930 0.9959 0.9935 0.9819 0.9990 0.9980 0.9986 0.9978 0.9990 0.9990 0.9990 0.9990 0.9976 0.9950 0.9992 0.9992 0.9996 0.9990 0.9990 0.9974

0.0070 0.0120 0.0016 0.0030 0.0055 0.0173 0.0004 0.0012 0.0010 0.0006 0.0006 0.0002 0.0006 0.0006 0.0020 0.0038 0.0006 0.0006 0.0002 0.0008 0.0006 0.0020

Dapp is the apparent diffusion coefficient determined from the linear fit of data curved due to contributions from intermolecular cross relaxation.

the (R)-BOH mixture. This result is expected if the curvature results from intermolecular cross relaxation. Because the analyte protons that are closest to MM atoms in the bound state experience intermolecular cross relaxation during the long diffusion time, the plots of ln(area) vs q for these atoms’ resonances show the greatest deviation from linearity. Analyte atoms that are further from the MM protons in the bound state do not experience intermolecular cross relaxation, and therefore, little to no deviation is expected. In addition, the curvature observed in the analyte-MM PFG-NMR experiments is not likely caused by baseline distortions or interference from intense MM peaks, as can be a problem in NOESY spectra, because only analyte aromatic resonances were analyzed and these peaks were well separated from the broad, intense MM peaks. Temperature variations, gradient nonlinearity, or other sources of instrumental error would also not be expected to affect one aromatic resonance more than the others because peaks of comparable signal to noise ratio were analyzed for each proton in the (R)BOH, (+)-TB, and (S)-propranolol experiments. The R2 values for the linear and polynomial fits of the analyte-MM PFG-NMR data also suggest that intermolecular cross relaxation was observed in these mixtures. Not surprisingly, in each analysis, the data were always better fit by a polynomial. However, the differences in the R2 values for the linear and nonlinear fits were always largest for the resonances with the largest κ values. For example, in the analysis of the (S)-propranolol-poly(SULV) results, resonance H7′ exhibited the smallest κ and the R2 values for the linear (0.9990) and polynomial (0.9996) fits are almost identical. Larger differences in the linear and polynomial R2 values are observed for (S)propranolol resonances H2′ and H8′, where the κ values are larger. Figure 5c shows residual plots for the linear and polynomial fits of the warfarin-AGP and (+)-TB-poly(SULV) diffusion data. The (+)-TB residuals correspond to the fits for resonance H7 (H1), which exhibited the largest κ value. The polynomial fit residuals for both the chiral analyte-MM and protein-ligand data sets are randomly distributed about zero. In contrast, the linear residuals in both analyses are negative at low q values,

pass through zero, reach a maximum, and then become negative again. This pattern was also observed for the linear fits of the resonances exhibiting the largest κ values in the (R)-BOH and (S)-propranolol-poly(SULV) data sets (data not shown). The observation that both the warfarin-AGP and analyte-MM PFG-NMR data sets exhibit the same residual pattern and thus similar deviations from linearity suggests that modest intermolecular cross relaxation effects were observed in the analytepoly(SULV) PFG-NMR experiments. It should be noted that all NMR gradient amplifiers have limited linearity and at high gradient amplitudes can generate nonlinear responses. This effect could lead to curvature in the linearized diffusion plots. In each mixture, the MM had the smallest diffusion coefficient; therefore, gradients up to 42.7 G/cm were needed to attenuate the MM signals by an order of magnitude. However, in the plots of the MM peak areas vs q, no curvature was observed and the R2 values for the linear fits all exceeded 0.999. Therefore, it is unlikely that gradient nonlinearity led to the curvature observed for the analyte resonances. As discussed above, the analyte or ligand protons that are in closest spatial proximity to protein or MM atoms are expected to have the largest κ values. The magnitude of κ is also dependent upon the association constant of the protein-ligand or analyte-MM complex, with higher K values corresponding to more curvature in the PFG-NMR data sets and thus larger κ values. The magnitude of the intermolecular cross relaxation detected in PFG-NMR is also dependent upon the motional properties of the bound analyte. For example, if an analyte is bound to a large protein or MM, then we would expect the bound analyte to have a long correlation time, τc, resulting in larger κ values from more effective intermolecular cross relaxation. In contrast, if an analyte binds to a smaller macromolecule, its correlation time is not slowed to the extent discussed above, and the κ value may be smaller, even if the binding constant is large and analyte protons are very close to the macromolecule atoms. To “factor out” the contribution to κ from the macromolecule size, the product κDb was calculated for each of the analyses performed here as well as from a literature PFG-NMR result.26 Db is the diffusion coefficient of

17368 J. Phys. Chem. B, Vol. 110, No. 35, 2006

Figure 6. Correlation between the free energy of binding and κDobsd for chiral analyte-MM mixtures (S)-propranolol-poly(SULV), (R)BOH-poly(SULV), and (+)-Troger’s base-poly(SULV) and ligandprotein solutions containing warfarin-AGP and L-tryptophan-human serum albumin (Trp-HSA).

the bound analyte, which is assumed to be equal to the macromolecule D value. By multiplying κ by Db, we differentially increase its value for analytes bound to smaller macromolecules and differentially decrease its value for ligands bound to large molecules. Thus, the contribution to κ from the bound analyte correlation time is partially factored out, and the product κDb should scale linearly with the analyte’s free energy of binding. Figure 6 is a plot of the free energy of binding versus the quantity κDb for the PFG-NMR analyses reported here and for the L-tryptophan-human serum albumin (Trp-HSA) analysis reported by Lucas et al.26 The free energy of binding for the AGP-(R)-warfarin complex was calculated from the association constant reported by Nakagawa et al.31 In each case, the largest κ value was used to calculate the product κDb. Examination of the plot in Figure 6 shows that the product κDb scales linearly with the free energy of binding, with the (+)-TB- and (R)BOH-MM results falling on the same trend line as the proteinligand analyses. This result suggests that the modest curvature observed in these mixtures is the result of intermolecular cross relaxation between the MM and the analyte. The plot also suggests that systems with association constants comparable to those of the MM-analyte complexes (ca. 400 M-1) yield κ values that are near the lower limit detectable with PFG-NMR. The κ value from the (S)-propranolol-poly(SULV) experiment does not follow the trend described above. Under the conditions of the PFG-NMR experiments, (S)-propranolol was cationic, while the other analytes were neutral. Therefore, propranolol may bind to the anionic MM by inserting into the polymer’s chiral pocket as well as through a nonstereoselective electrostatic interaction with the negatively charged micelle. The former, but not the latter, binding site will lead to intermolecular cross relaxation and contribute to κ. As a result, (S)-propranolol has a large association constant, but still exhibits a κ value comparable to those of the other analytes, and the data point from (S)-propranolol lies below the trend line in Figure 6. The group binding maps generated with the intermolecular cross relaxation diffusion experiment can also be compared to the maps generated from the NOE difference and NOESY analyses. For (+)-TB and (S)-propranolol, we see that the PFGNMR map most closely resembles the NOE difference and NOESY maps to the valine HR. In the (+)-TB analysis, both

Morris et al. PFG-NMR and the NOESY valine HR maps have H7 (H1) at 100% and H9 (H3) with the smallest percentage. Likewise for (S)-propranolol, both the NOESY valine HR and PFG-NMR analyses set 100% at the analyte proton H2′, which is in close proximity to the compound’s chiral side chain. The PFG-NMR map for (R)-BOH sets 100% at H4′ and shows high percentages at H5′, H6′, and H3′ and the lowest percentage at H8′. Neither NOESY nor NOE difference mapping experiments rank the percentages in this exact order. Therefore, it is probable that the intermolecular cross relaxation diffusion map generated for (R)-BOH represents an average or superposition of interactions with both the hydrocarbon chain and MM headgroup protons. The (R)-BOH map points out one potential drawback of binding interaction mapping with the intermolecular cross relaxation diffusion method. Since the NOE effects are not generated by saturating or inverting the populations of specific macromolecule protons, the cross relaxation detected and the maps generated with this experiment may not distinguish between the analyte interactions at the MM chiral center or at another specific binding site of interest. In contrast, NOESY and NOE difference methods map analyte interactions at a specific binding site of interest. The mapping techniques employed here are also applicable to other EKC chiral selectors such as cylcodextrins or chiral surfactant micelles.16 As with the analyte-MM analyses, intermolecular cross relaxation in PFG-NMR will likely be relatively small in most other analyte-modifier applications. However, diffusion experiments can still be used to measure enantiomeric association constants and to quantify each enantiomer’s free energy of binding. Interaction mapping with NOE difference, NOESY, or ROESY experiments, however, should allow the mechanism of chiral recognition and the orientation of the bound analyte to be elucidated in applications with cyclodextrin or chiral micelle modifiers. This information will help guide the optimization of separation conditions or the design of new modifier systems.16 Conclusions Pulsed field gradient NMR diffusion experiments were used to measure MM association constants for the enantiomers of four chiral analytes. With the analytes BNP, BOH, and Troger’s base, these association constants were found to be consistent with EKC elution order in that the analyte enantiomer with the longer EKC retention time was found to have the larger MM binding constant. Group epitope mapping experiments originally developed to investigate protein-ligand binding were also applied to the study of chiral recognition by MM. Maps generated by NOESY and NOE difference experiments show reasonable agreement and suggest similar mechanisms for analyte interaction with poly(SULV), providing a rationale for the use of this MM as a selector in chiral separations. Subtle NOE effects were also observed in PFG-NMR experiments with MM-chiral analyte mixtures. These effects produced differential curvature in ln(area) vs q plots. The magnitude of intermolecular cross relaxation in NMR diffusion experiments was also observed to scale linearly with the free energy of binding. Acknowledgment. C.K.L. gratefully acknowledges financial support from the National Science Foundation, Grant CHE 0213407, and I.M.W. acknowledges support from the National Institutes of Health. B.A.B. acknowledges support from a National Institutes of Health training grant on Pharmaceutical Aspects of Biotechnology, GM-08359. Acknowledgment is also made to the donors of the Petroleum Research Fund, adminis-

Interactions of Chiral Molecules with MMs tered by the American Chemical Society, for partial support of this research through Grants 42859-AC4 to C.K.L. and 39394B4 to K.F.M. References and Notes (1) Heller, M.; Kessler, H. Pure Appl. Chem. 2001, 73, 1429-1436. (2) Wang, J.; Warner, I. M. Anal. Chem. 1994, 66, 3773-3776. (3) Billiot, E.; Macossay, J.; Thibodeauz, S.; Shamsi, S. A.; Warner, I. M. Anal. Chem. 1998, 70, 1375-1381. (4) Billiot, E.; Warner, I. M. Anal. Chem. 2000, 78, 1740-1748. (5) Billiot, F. H.; Billiot, E. J.; Warner, I. M. J. Chromatogr., A 2001, 922, 329-338. (6) Billiot, E. J.; Agbaria, R. A.; Thibodeaux, S.; Shamsi, S.; Warner, I. M. Anal. Chem. 1999, 71, 1252-1256. (7) Billiot, E. J.; Thibodeaux, S.; Shamsi, S.; Warner, I. M. Anal. Chem. 1999, 71, 4044-049. (8) Rugutt, J. K., Billiot, E. J.; Warner, I. M. Langmiur 2000, 16, 30223029. (9) Yarabe, H. H.; Rugutt, J. K.; McCarroll, M. E.; Warner, I. M. Electrophoresis 2000, 21, 2025-2032. (10) Shamsi, S. A.; Palmer, C. P.; Warner, I. M. Anal. Chem. 2001, 73, 140A-149A. (11) Billiot, F. H.; McCarroll, M. E.; Billiot, E. J.; Warner, I. M. Electrophoresis 2004, 25, 753-757. (12) Harrell, C. W.; McCarroll, M. E.; Morris, K. F.; Billiot, E. J.; Warner, I. M. Langmuir 2003, 19, 10684-10691. (13) Mwongela, S.; Akbay, C.; Zhu, X.; Collins, S.; Warner, I. M. Electrophoresis 2003, 24, 2940-2947. (14) Billiot, F. H.; Billiot, E. J.; Warner, I. M. J. Chromatogr., A 2002, 950, 233-239.

J. Phys. Chem. B, Vol. 110, No. 35, 2006 17369 (15) Thibodeaux, S. J.; Billiot, E. J.; Warner, I. M. J. Chromatogr., A 2002, 966, 179-186. (16) Morris, K. F.; Froberg, A. L.; Becker, B. A.; Almeida, V. K.; Tarus, J.; Larive, C. K. Anal. Chem. 2005, 77, 254A-263A. (17) Meyer, B.; Weimar, T.; Peters, T. Eur. J. Biochem. 1997, 246, 705709. (18) Mayer, M.; Meyer, B. Angew. Chem., Int. Ed. 1999, 38, 17841788. (19) Meinecke, R.; Meyer, B. J. Med. Chem. 2001, 44, 3059-3065. (20) Mayer, M.; Meyer, B. J. Am. Chem. Soc. 2001, 123, 6108-6117. (21) Peikert, C.; Seeger, K.; Bhat, R. K.; Berger, S. Org. Biomol. Chem. 2004, 2, 1777-1781. (22) Johnson, M. A,; Pinto, B. M. Bioorg. Med. Chem. 2004, 12, 295300. (23) Murata, T.; Hemmi, H.; Nakajimi, M.; Yoshida, M.; Yamaguchi, I. Biochem. Biophys. Res. Commun. 2003, 307, 498-502. (24) Chatterjee, C.; Majumder, B.; Mukhopadhyay, C. J. Phys. Chem. 2004, 108, 7430-7436. (25) Yan, J.; Kline, A. D.; Mo, H.; Zartler, E. R.; Shapiro, M. J. J. Am. Chem. Soc. 2002, 124, 9984-9985. (26) Lucas, L. H.; Yan, J.; Larive, C. K.; Zartler, E. R.; Shapiro, M. J. Anal. Chem. 2003, 75, 627-634. (27) Lucas, L. H.; Larive, C. K. Concepts Magn. Reson., A 2004, 20A, 24-41. (28) Wu; D.; Chen. A.; Johnson, C. S., Jr. J. Magn. Reson., Ser. A 1995, 115, 260-264. (29) Uedaira H.; Uedaira H. J. Phys. Chem. 1970, 74, 2211. (30) Valle, B. C.; Morris, K. F.; Fletcher, K. A.; Fernand, V.; Sword, D. M.; Eldridge, S.; Larive, C. K.; Warner, I. M. Langmuir, submitted for publication. (31) Nakagawa, T.; Kishino, S.; Itoh, M.; Sugawara, M.; Miyazaki, K. Br. J. Clin. Pharmacol. 2003, 56, 664-669.