Acetylation of β-Cyclodextrin Surface-Functionalized Cellulose

enantioselectivity in the range of 1.26-1.33 depending on the acetylation time. The improvement in enantioselectivity after acetylation was mainly att...
0 downloads 0 Views 468KB Size
Download PDF
12990

Langmuir 2007, 23, 12990-12996

Acetylation of β-Cyclodextrin Surface-Functionalized Cellulose Dialysis Membranes with Enhanced Chiral Separation Youchang Xiao,† Hui Miang Lim,† Tai Shung Chung,*†,‡ and Raj Rajagopalan† Department of Chemical and Biomolecular Engineering and Singapore-MIT Alliance, National UniVersity of Singapore, 10 Kent Ridge Crescent, Singapore 119260 ReceiVed August 27, 2007. In Final Form: September 20, 2007 The enhanced enantiomeric separation of racemic phenylalanine solution has been demonstrated by the membranebased chiral resolution method using an acetylated β-cyclodextrin-immobilized cellulose dialysis membrane. β-Cyclodextrin (CD) was first immobilized onto the surface of commercial cellulose dialysis membranes, followed by the acetylation reaction through the treatment of the membranes with acetic anhydride to form the chiral selective acetylated β-cyclodextrin-immobilized cellulose dialysis membrane. The acetylated CD-immobilized membrane exhibits enantioselectivity in the range of 1.26-1.33 depending on the acetylation time. The improvement in enantioselectivity after acetylation was mainly attributed to the better discrimination ability of acetylated CD and the decrease in membrane pore size. Molecular modeling simulations indicate that the acetylation of hydroxyl groups would result in a CD conformation with torus distortions and would create higher steric hindrance for penetrants. As a result, compared to the original CD, the acetylated CD may have less effective binding but better discrimination of enantiomers. The energy drop is only 3 kcal/mol between different enantiomers before and after the binding of phenylalanine with an unmodified CD. The energy drop increases to 10 kcal/mol if acetylated CD is employed as the chiral selector, showing stronger characteristics for chiral selection.

1. Introduction The demands of the pharmaceutical industry provide the impetus to develop suitable techniques for the resolution of enantiomers because the majority of active pharmaceutical ingredients (API) are chiral in nature.1 Currently, crystallization is the most widely large-scale process for enantiomeric resolution.2 However, this process requires large amounts of mother liquor and leads to high losses of valuable product. A variety of other techniques exist for chiral separation, including chromatography and electrophoresis, but most of them are on an analytical scale and allow only a small amount of materials to be separated per run; their scale-up is expensive and not energy- and costefficient.3,4 Membrane-based chiral separation shows strong competition, and it may potentially overcome the shortcomings of the aforementioned processes. In addition, membrane separation technology can be performed continuously under ambient conditions, and process scale up is relatively easy.5-7 Membranes for chiral separation can be divided into liquid and solid membranes.8 In liquid membrane systems, optical resolution was carried out by either bulk, supported, or emulsion liquid membranes. High enantioselectivity but a short membrane usage life may be obtained.9,10 To overcome the drawbacks of liquid membranes, solid membranes such as chiral polymers, * Corresponding author. E-mail: [email protected]. Fax: (65) 6779 1936. † Department of Chemical and Biomolecular Engineering. ‡ Singapore-MIT Alliance. (1) Maier, N. M.; Franco, P.; Linder, W. J. Chromatogr., A 2001, 3, 906. (2) Li, J. Z.; Grant, D. J. W. J. Pharm. Sci. 1997, 86, 1073. (3) Ravelet, C.; Peyrin, E. J. Sep. Sci. 2006, 29, 1322. (4) Wistuba, D.; Schurig, V. J. Chromatogr., A 2000, 875, 255. (5) Baker, R. W. Membrane Separation Systems: Recent DeVelopments and Future Directions; Noyes Data Corp.: Park Ridge, NJ, 1991. (6) Charcosset, C. Biotechnol. AdV. 2006, 24, 482. (7) Jirage, K. B.; Martin, C. R. Trends Biotechnol. 1999, 17, 197. (8) Van der Ent, E. M.; van’t Riet, K.; Keurentjes, J. T. F.; van der Padt, A. J. Membr. Sci. 2001, 185, 207. (9) Mandal, D. K.; Guha, A. K.; Sirkar, K. K. J. Membr. Sci. 1998, 144, 13. (10) Armstrong, D. W.; Jin, H. L. Anal. Chem. 1987, 59, 2237.

molecularly imprinted membranes, and chiral selector immobilized membranes have received much greater attention. These solid membranes are fairly stable; therefore, a durable enantiomer separation process may be achievable. On the basis of the separation mechanism, solid membranes for chiral separation might be categorized into two types: diffusion-selective membranes and sorption-selective membranes. A diffusion-selective membrane is defined as a membrane without specific foreign chiral selectors for chiral interaction but made of a polymer with intrinsic chiral separation characteristics. The membrane can be either in a composite where the polymer is coated on a microporous layer or in a self-supporting configuration.11 The main disadvantage of this type of membranes is that it has high selectivity but very low permeability or vice versa. Sorptionselective membranes have their own characteristics: no selective permeation can be observed, despite the fact that they show selective sorption, which is achieved by embedding or immobilizing chiral selectors in a polymer matrix or on the membrane surface. In most cases, these selectors are known from analytical separation methods. Examples are cyclodextrins,12,13 crown ether derivatives,14 and some biomaterials (proteins, DNA, and antibodies).15-18 Additionally, molecular imprinted polymers also can be regarded as sorption-selective membranes because the formed cavities during the imprinting procedure can form one-to-one molecular complexes with preferred enantiomers.19 Recently, immobilizing chiral carriers into solid membranes or upon the surface of membranes has also been reported with (11) Aoki, T.; Kaneko, T. Polym. J. 2005, 37, 717. (12) Krieg, H. M.; Breytenbach, J. C.; Keizer, K. J. Membr. Sci. 2000, 164, 177. (13) Xiao, Y. C.; Chung, T. S. J. Membr. Sci. 2007, 290, 78. (14) Kakuchi, T.; Takaoka, T.; Yokota, K. Polym. J. 1990, 22, 199. (15) Higuchi, A.; Higuchi, Y.; Furuta, K.; Yoon, B. O.; Hara, M.; Maniwa, S.; Saitoh, M.; Sanui, K. J. Membr. Sci. 2003, 221, 207. (16) Higuchi, A.; Yomogita, H.; Yoon, B. O.; Kojima, T.; Hara, M.; Maniwa, S.; Saitoh, M. J. Memb. Sci. 2002, 205, 203. (17) Lakshmi, B. B.; Martin, C. R. Nature 1997, 388, 758. (18) Higuchi, A.; Hayashi, A.; Kanda, N.; Sanui, K.; Kitamura, H. J. Mol. Struct. 2005, 739, 145. (19) Ramstrom, O.; Ansell, R. J. Chirality 1998, 10, 195.

10.1021/la7026384 CCC: $37.00 © 2007 American Chemical Society Published on Web 11/17/2007

Acetylation of β-Cyclodextrin Dialysis Membranes

higher membrane durability and comparable chiral separation performance. We have reported the chiral separation of racemic amino acid tryptophan using β-cyclodextrin-immobilized membranes prepared from commercial cellulose dialysis membranes.13 Native cyclodextrins (CD) are cyclic oligosaccharides consisting of six to eight D-(+)-glucopyranose units, providing three-point interactions for the chiral recognition of various organic molecules: hydrophobic interaction in the CD cavity and two hydrogen bonds with the hydroxyl groups at the opening of the CD.20 Chiral separation of a racemic tryptophan solution using an immobilized β-cyclodextrin (CD) membrane, having a molecular weight cutoff (MWCO) of 1000, resulted in more rejection to L-tryptophan than D-tryptophan and exhibited an enantioselectivity of around 1.10. Compared with other chiral selector-immobilized membranes, CD-functionalized membranes are lower in cost and have wider applicability and higher tolerance in various environments. However, native cyclodextrins have limited chiral recognition ability. CD’s chemically modified derivatives are an important topic of research and have been used for various purposes, such as high performance liquid chromatography (HPLC) and capillary electrophoresis (CE). The objective of this work is to enhance chiral separation performance by the chemical modification of native cyclodextrins. It is believed that certain chemical modification of native cyclodextrins is able to change the depth of the cyclodextrin cavity, the hydrogen-bonding ability, and other various physical properties. The literature indicates that the enantiomeric separation of most amino acids changed after acetylation of the cyclodextrinbonded stationary phases in HPLC.20 However, heptakis(2,3di-O-acetyl)-β-cyclodextrin and heptakis(6-mono-O-acetyl)-βcyclodextrin have been applied to chiral separations by HPLC and capillary electrophoresis.21-26 Acetylated β-cyclodextrins are expected to exhibit different enantioselectivity in comparison with native cyclodextrins because the acetyl groups are bulky and have the ability to form hydrogen bonds. To the best of our knowledge, no reports exist on the acetylation of immobilized β-cyclodextrin (CD) onto the surface of membranes for a chiral separation process. In this article, we report the chiral separation of phenylalanine amino acid using acetylated CD-immobilized membranes with the aid of a dialysis permeation cell setup, where the concentration gradient is the only driving force. We specifically focus on whether the increase in the acetylation reaction time of acetic anhydride will have an effect on the enantioselectivity and the permeation flux of enantiomers through the membranes. 2. Experimental Section 2.1. Materials. The base membranes used for the chemical modification were cellulose dialysis membranes, which are available commercially (Spectra/Pro7, MWCO ) 1000, membrane thickness ) 40 µm, Spectrum Medical Industries, Inc). 6-Monodeoxy-6monoamino-β-cyclodextrin hydrochloride (C42H72O34NCl) (N-CD) from Cyclolab, sodium periodate, 1,4-dioxane (99.8% anhydrous), and sodium cyanoborohydride from Aldrich were used as received. (20) Schneiderman, E.; Stalcup, A. M. J. Chromatogr., B 2000, 745, 83. (21) Chankvetadze, B.; Lomsadze, K.; Burjanadze, N.; Breitkreutz, J.; Pintore, G.; Chessa, M.; Bergander, K.; Blaschke, G. Electrophoresis 2003, 24, 1083. (22) Tanaka, M.; Shono, T.; Zhu, D. Q.; Kawaguchi, Y. J. Chromatogr. 1989, 469, 429. (23) Branch, S. K.; Holzgrabe, U.; Jefferies, T. M.; Mallwitz, H.; Matchett, M. W. J. Pharm. Biomed. Anal. 1994, 12, 1507. (24) Matchett, M. W.; Branch, S. K.; Jefferies, T. M. J. Chromatogr., A 1995, 705, 351. (25) Branch, S. K.; Holzgrabe, U.; Jefferies, T. M.; Mallwitz, H.; Oxley, F. J. R. J. Chromatogr., A 1997, 758, 277. (26) Miura, M.; Kawamoto, K.; Funazo, K.; Tanaka, M. Anal. Chim. Acta 1998, 373, 47.

Langmuir, Vol. 23, No. 26, 2007 12991 Other reagents such as anhydrous disodium hydrogen phosphate and monosodium dihydrogen phosphate monohydrate from Merck, acetic anhydride from Fluka, 3-picoline from Acros and DLphenylalanine amino acid from Alfa Aesar were reagent grade and were used without further purification. 2.2. Preparation of Acetylated CD-Immobilized Cellulose Dialysis Membranes. The acetylated CD-immobilized cellulose dialysis membranes can be prepared in two steps: preparation of immobilized CD cellulose dialysis membranes, followed by acetylation of the CD-immobilized cellulose dialysis membranes. The immobilized CD cellulose dialysis membranes were first prepared according to the method reported in our previous study.13 A cellulose dialysis membrane with a MWCO of 1000 was first soaked in deionized water for 0.5 h at room temperature to remove the residual storage chemical, sodium azide, which is used to prevent bacteria propagation in the membrane. The membrane was then soaked in 50 mL of a 1 N NaIO4 solution for 2 h at room temperature. After being washed with deionized water three times and phosphate buffer solution (500 mM, pH 6.0) three times, the oxidized membrane was soaked in 50 mL of phosphate buffer solution (500 mM, pH 6.0) containing 2 mM N-CD for another 2 h at room temperature. The membrane was then soaked in 50 mL of phosphate buffer solution (500 mM, pH 6.0) containing 10 mM NaCNBH3 for another 2 h at room temperature. Finally, the immobilized CD membrane was prepared. For the acetylation reaction, the immobilized CD membrane was soaked in anhydrous 1,4-dioxane for 1 h at room temperature three times to get rid of residual water in the membrane. Then the membrane was immersed in a mixture of 20 mL of anhydrous 1,4-dioxane, 20 mL of acetic anhydride, and 20 mL of 3-picoline and stirred for 0.5, 1, and 2 h at 70 °C, respectively. The resulting acetylated CDimmobilized membrane was washed with deionized water three times and preserved in deionized water for testing. The above procedure for the preparation of acetylated CD-immobilized cellulose dialysis membranes is illustrated in Figure 1. 2.3. Surface Characterization of the Modified Membranes. An X-ray photoelectron spectrometer was used to measure the element ratio and monitor the chemical reaction on the cellulose membrane surface. The XPS measurements were carried out with an AXIS HSi spectrometer (Kratos Analytical Ltd., England) using a monochromatized Al KR X-ray source (1486.6 eV photons) at a constant dwell time of 100 ms and a pass energy of 40 eV. The anode voltage and anode current were 15 kV and 10 mA, respectively. The pressure in the analysis chamber was maintained at e5.0 × 10-8 Torr during each measurement. All core-level spectra were obtained at a photoelectron takeoff angle of 90° with respect to the sample surface, and the X-ray penetration depth was about 5-7 nm for polymer materials. The error range for the XPS element analysis was less than 5%. FTIR-ATR (attenuated total reflection) measurements were carried out using a Perkin-Elmer FTIR microscope at 8 cm-1 resolution over the 500-2200 cm-1 range. Each sample was scanned 20 times. 2.4. Chiral Separation through Acetylated CD-Immobilized Membranes. The dialysis membranes were placed in a permeation cell similar to the setup used in our previous study, as shown in Figure 2. This permeation cell is designed to evaluate the permeation of a racemic 0.1 mM phenylalanine solution. The cell has two Teflon chambers, each holding a volume of 40 mL of solution. One compartment is charged with chiral feed solution and the other compartment holds the strip solution. One liter of 0.1 mM racemic DL-phenylalanine solution was prepared as the feed phase that was circulated between a reservoir and the cell. However, the strip phase was 40 mL of pure deionized water. The membrane to be tested was placed in between the two compartments at the center of the permeation cell, acting as a vertical flat membrane system. The effective membrane surface area was 4.9 cm2. Both of the aqueous phases were mechanically stirred with Teflon impellers that were connected to an overhead mixer (CAT R18, M. Zipperer GmbH, Staufrn, Germany) at 200 rpm. Samples were taken regularly at 0.5 h intervals from the strip phase in 100 µL and analyzed by HPLC

12992 Langmuir, Vol. 23, No. 26, 2007

Xiao et al.

Figure 1. Preparation of acetylated CD-immobilized cellulose dialysis membranes.

Figure 2. Schematic of the dialysis system to test the membranes. (Agilent Co.) with a Chirobiotic T column (Astec Inc.). The analytical conditions were as follows: a mobile phase composition of ethanol/ water (80/20), flow rate of 0.8 mL/min, column temperature of 298 K, and UV detection at 210 nm. The permeation test lasted for 300 min to ensure that the concentration difference between the two phases could be assumed to be constant. 2.5. Computational Simulation of the Combination between CDs and Enantiomers. All molecular modeling studies were performed with the Material Studio 4.1 software pack published by Accelrys Inc. The initial structures of CDs, amino acids, and complexes were constructed and visualized using the builder module and then were optimized by a molecular mechanics technique to achieve the respectively minimized energy using the minimizer in the discover module. The compass force field and 100 000 iterations were used for all calculations.

3. Results and Discussion 3.1. Characterization of Acetylated CD-Immobilized Membranes. The support membranes used in this work are commercial regenerated cellulose membranes with a MWCO of only 1000 Da. The pore size is smaller than 1 nm, which is much smaller than the size of CD. Therefore, we believe that CD modification occurs only on the surface of the support membranes. Element analyses on the surface of the modified cellulose dialysis membranes (MWCO ) 1000) were conducted by XPS to monitor

the modification process, and the results are shown in Table 1. Because both cellulose and CD consist of glucose units, the number of carbon and oxygen atoms in membranes before modification is almost identical to that after modification. The trace of the nitrogen atom of the pristine cellulose membrane comes from the residual storage chemical sodium azide, which prevents membranes from breeding bacteria. The N content is reduced after the oxidation step, indicating that the sodium azide was washed out of the membrane. However, the number of nitrogen atoms, which originated from N-CD, increases after the immobilization step, suggesting the immobilization of CD on the surface of cellulose membranes. The acetylation process decreases the concentration of nitrogen on the membrane surface. The main reason for the decrease in N may be that the N atom concentration was diluted by the acetyl groups. The ratio of carbon to oxygen atoms also does not show an obvious change after acetylation because of the equal carbon/oxygen ratio of glucose units and acetyl groups. The spectra of samples with longer modification are not much different. This may be due to the fact that the surface modification reaction is almost saturated in the first half hour. Further acetylation mainly occurs in the support layer. Therefore, only the spectrum of samples with 0.5 h of acetylation was shown here for the confirmation of surface acetylation.

Acetylation of β-Cyclodextrin Dialysis Membranes

Langmuir, Vol. 23, No. 26, 2007 12993

Table 1. Surface Element Analysis by XPS for Various Cellulose Membranesa

a

N atomic content of NH2-CD: 1.3%

The chemical reaction during the whole modification process can be proven by the evolution of the respective O 1s core-level spectra of the pristine cellulose membrane, oxidized cellulose membrane, CD-immobilized membrane, and acetylated CDimmobilized membrane. The O 1s core-level spectra of the pristine cellulose membrane can be curve-fitted to two peak components, having a bond energy of 533 eV for the C-O-C species and 532.5 eV for the C-O-H species. After being oxidized by NaIO4, the carbon-carbon bond of the vicinal diol unit is broken, and two carbonyl groups resulted. Consequently, the O 1s core-level spectra of the oxidized cellulose membrane show a new peak at a bond energy of around 531 eV, contributing to the newly formed carbonyl groups. The resulting carbonyl groups can react with the primary amine groups of N-CD at the optimum pH, usually about 5 to 6, at which the reaction rate is a maximum. The vanishing of the characteristic peak of the carbonyl group in the O 1s core-level spectra, after the oxidized membrane is soaked in N-CD solution, proves the reaction between the oxidized cellulose membrane and N-CD. After the last 0.5 h in the acetylation stage, the O 1s core-level spectra indicate that new carbonyl groups are induced and C-O-H groups are consumed. Surface characterization of the modified membranes was also carried out using FTIR-ATR spectrometry. The results are shown in Figure 3. The immobilized CD cellulose membrane and the pristine cellulose membrane have similar spectra. The results are expected because both the cellulose and CD consist of glucose units. However, the acetylated CD-immobilized membrane shows peaks detected at wavenumbers of 1230 and 1730 cm-1. The 1230 cm-1 wavenumber corresponds to the asymmetric stretching vibration of C-O for aliphatic esters, and 1730 cm-1 corresponds to the stretching vibration of CdO for the ester group. The weak signal detected at 1370 cm-1 shows the presence of the acetate group. Hence, from the FTIR spectra, it can be confirmed that the acetylation reaction modified the membrane surface. However, we still cannot quantitatively know how much CD was immobilized onto the surface by direct measurements. The surface concentration of CD will definitely affect the chiral separation performance of the resultant membranes. This will be the focus of future work.

3.2. Chiral Separation by CD-Immobilized Membranes. The study of separating a racemic mixture of phenylalanine was conducted by the dialysis system using three types of MWCO 1000 membranes: (1) pristine cellulose dialysis membranes, (2) CD-immobilized membranes, and (3) acetylated CD-immobilized membranes. The concentration of the enantiomers in the strip solution chamber increased linearly as a function of time during the permeation test period of 300 min. In this experiment, the racemic phenylalanine feed solution in the feed chamber was 0.1 mM and was assumed to be constant during the experimental period of 300 min. The flux J (mol/cm2 h) can be calculated from the slope using

J)

∆CV ∆tA

(1)

where ∆C is the change in concentration, ∆t is the permeation time, V is the strip volume (40 cm3), and A is the effective membrane area (4.9 cm2). The permeability coefficient P (cm2/ s) is given by

P)

Jd C F - CS

(2)

where d is the membrane thickness and CF - CS is the concentration difference between the feed-phase chamber and the strip-phase chamber. After 300 min of operation, the concentration of the strip volume is only 0.01 mM, so the final concentration of the feed solution (0.1 mM) could be considered to be almost constant. The enantioselectivity R is the ratio of the permeability coefficients of D-phenylalanine to the L-phenylalanine (PD/PL).

R)

PD PL

(3)

A summary of the resultant permeabilities of the two isomers and enantioselectivity is given in Table 2. From Table 2, the pristine cellulose dialysis membrane does not show chiral selectivity (R ) 0.99) for the racemic mixture of phenylalanine.

12994 Langmuir, Vol. 23, No. 26, 2007

Xiao et al.

Figure 3. FTIR-ATR spectra of various membranes. Table 2. Permeability and Enantioselectivity in the Chiral Separation of Racemic Phenylalanine through Various Membranes membrane

PL (cm2/s)

pristine cellulose membrane immobilized CD membrane acetylated CD-immobilized membrane (0.5 h) immobilized CD membrane* (for tryptophan)

(2.49 ( 0.12) × (2.20 ( 0.11) × 10-7 (1.22 ( 0.06) × 10-7 (1.51 ( 0.08) × 10-7

This result is expected because the chiral selectors used to separate the two enantiomers are absent in the pristine cellulose dialysis membrane. After the membrane was immobilized with CD, the permeation fluxes of the two enantiomers decrease but at different rates. The immobilized CD membrane shows more resistance to Lphenylalanine than to D-phenylalanine and exhibits an enantioselectivity of 1.11. This can be explained by the model of the thin chiral solution layer near the surface of the membrane as proposed in our previous study.13 Stronger interaction between the L-phenylalanine and CD results in more L-phenylalanine complexed by the CD upon the membrane surface. Accordingly, there is a lower concentration of free and mobile L-phenylalanine than free and mobile D-phenylalanine in the chiral solution layer near the surface of the membrane. Although the diffusion coefficients of both enantiomers through the supported cellulose membrane are the same, the driving force of D-phenylalanine is greater than that of L-phenylalanine. Consequently, the apparent permeability of D-phenylalanine is faster than that of Lphenylalanine. Therefore, enantioselectivity is observed in the immobilized CD membrane. The results for the chiral separation of racemic phenylalanine are consistent with the previous study on the chiral separation of racemic tryptophan using the same immobilized CD membranes. In the previous study on racemic tryptophan, the immobilized CD cellulose dialysis membranes (MWCO ) 1000) had permeability coefficients of PD equal to 1.66 × 10-7 cm2/s and PL equal to 1.51 × 10-7 cm2/s, and an enantioselectivity of 1.10. In the present work, the immobilized CD cellulose dialysis membranes (MWCO ) 1000) have higher permeability coef-

10-7

PD (cm2/s)

enantioselectivity, R

(2.47 ( 0.12) × 10-7 (2.44 ( 0.12) × 10-7 (1.53 ( 0.08) × 10-7 (1.66 ( 0.09) × 10-7

0.99 1.11 1.26 1.10

Table 3. Permeability and Enantioselectivity in the Chiral Separation of Racemic Phenylalanine through Acetylated CD-Immobilized Membranes with Varying Acetylation Reaction Time length of acetylation reaction time (h) 0.5 1.0 2.0

PL (cm2/s)

PD (cm2/s)

(1.22 ( 0.06) × (1.53 ( 0.08) × 10-7 (1.11 ( 0.06) × 10-7 (1.44 ( 0.07) × 10-7 (9.62 ( 0.48) × 10-8 (1.28 ( 0.06) × 10-7 10-7

enantioselectivity, R 1.26 1.30 1.33

ficients for both isomers of phenylalanine (PD ) 2.44 × 10-7 cm2/s and PL ) 2.20 × 10-7 cm2/s), but the enantioselectivity was 1.10, which was the same as in the previous study. Higher permeability coefficients for both isomers of racemic phenylalanine are expected because the molecular structure of phenylalanine is smaller than that of tryptophan. Hence the D- and L-phenylalanine molecules would be able to diffuse through the pores of the immobilized CD membranes faster than would the D- and L-tryptophan molecules. 3.3. Chiral Separation by Acetylated CD-Immobilized Membranes. As shown in Table 2, after the membrane has been immobilized with CD and followed by the acetylation reaction, the permeation flux of the two phenylalanine enantiomers decreases further but with different decreasing rates as compared to that of the pristine cellulose membrane. The acetylated CDimmobilized membrane shows more rejection to L-phenylalanine than D-phenylalanine and exhibits an enantioselectivity of 1.26. Furthermore, the effect of the acetylation reaction time on the permeation flux and enantioselectivity was investigated. As shown

Acetylation of β-Cyclodextrin Dialysis Membranes

Figure 4. Simulation results of native CD and

Langmuir, Vol. 23, No. 26, 2007 12995

DL-phenylalanine

Figure 5. Simulation results of acetylated CD and

combinations.

DL-phenylalanine

in Table 3, when the acetylation reaction time increases from 0.5 to 1 and then to 2 h, the acetylated CD-immobilized cellulose dialysis membranes show an incremental drop in the permeation flux of the two isomers. The acetylated CD-immobilized membrane that had been acetylated for 2 h shows more resistance to L-phenylalanine than D-phenylalanine and exhibits an enantioselectivity of 1.33 which is higher than the enantioselectivity

combinations.

for the other two membranes that have been acetylated for 0.5 and 1 h, respectively. Two possible reasons can help explain the observation of higher enantioselectivity obtained for the acetylated CD-immobilized membrane. One is the ester groups grafted on CD after acetylation, which may provide additional steric hindrance to the two isomers near the chiral center, thus enhancing chiral recognition. This

12996 Langmuir, Vol. 23, No. 26, 2007

Xiao et al.

Figure 6. Possible chiral separation model based on modified porous membranes.

phenomenon may be explained by simple molecular models of native cyclodextrin and its derivative. The structures of native cyclodextrin and acetylated cyclodextrin were drawn using the Discover model in the Accelrys Material Studio software package. The left-hand sides of Figures 4 and 5 show the top views of the structures with minimized energy for the cavity of each cyclodextrin with the most stable structure. A comparison of these Figures shows that the acetylation of hydroxyl groups can give a CD conformation with torus distortions. Steric hindrance apparently expands the substituted rim, and a distortion of the circular shape of the cavity occurs as well. This distortion could prevent a good fit of the penetrant in the cavity. As a result, the acetylated CD may reflect this less effective binding. To verify this assumption, the combinations of CD and phenylalanine (abbreviated as Pen) have also been drawn with minimized energy, as shown on the right-hand sides of Figures 4 and 5. By subtracting the minimized energy of CD and phenylalanine in the initial stage (i.e., both of them are separated) to that under the final stage (i.e., a combined state), we can estimate the energy drop during the binding of phenylalanine with each cyclodextrin. It was found that the energy drop becomes smaller from 60 kcal/ mol to around 30 kcal/mol when native CD was acetylated, indicating weaker binding between the acetylated CD and phenylalanine. However, the distorted cavity may cause the binding of a penetrant to depend more critically on the correct molecular geometry, and thus better discrimination of enantiomers would result. When comparing the energy drop between different enantiomers before and after the binding of phenylalanine with CD, we can determine the chiral discrimination ability of different selectors. If CD is used as the chiral selector, then the energydrop difference is only 3 kcal/mol, but after acetylation, the energy-drop difference increases to 10 kcal/mol. This data agrees with the above hypothesis that the acetylated CD shows stronger chiral selectivity. The other reason is the reduction in pore size of the membrane. After the acetylation reaction, the pores of the membrane become smaller as the OH groups on the cellulose dialysis membrane react with the acetyl groups to form esters. This causes additional steric hindrance and reduces the pore size of the membrane. Hence, the resistance for the two isomers to permeate through the membrane increases, and the permeation flux of the two isomers decreases further, as observed in Tables 2 and 3. Similar to our previous study,13 a thin chiral layer model could be used to explain this phenomenon, as shown in Figure 6. The formation of a chiral layer is expected at the membrane-solution interface. If the thickness of the chiral layer is larger than the pore radius of the membrane, then the driving force for amino acid diffusion

across the membrane is completely contributed from the chiral layer. However, if the membrane’s pore size is increased and becomes larger than the chiral layer thickness, then the racemic solution will be transported though the pore. As a consequence, the enantioselectivity increases with a decrease in membrane pore size. Moreover, the CD chiral layer does not show resistance to isomers compared to the resistance from the porous support. In other words, the support membranes dominate the isomer transport rate. The CD chiral layers change only the surface concentration of isomers because of the different bonding constants to isomers. Different surface concentrations induce different driving forces. Therefore, the modified membranes show some enantiomer selectivity. When we increase the resistance of support membranes (decreased pore size), the isomers take longer to pass through the CD chiral layer. If a longer duration is provided for the isomer exchange between CD and solution, then the selectivity will be enhanced and will move closer to the theoretical selectivity, which is the ratio of the binding constants of CD and different isomers. Therefore, the enantioselectivity is expected to be higher.

4. Conclusions In this study, commercial cellulose dialysis membranes were first impregnated with β-cyclodextrin, followed by the acetylation reaction using acetic anhydride to form chiral selective acetylated β-cyclodextrin-immobilized cellulose dialysis membranes. Chiral separation of racemic phenylalanine using acetylated CDimmobilized cellulose dialysis membranes was shown to be more favorable. As the reaction time for acetylation increases, the enantioselectivity of the acetylated CD-immobilized cellulose dialysis membranes increases, but the permeation flux of D- and L-phenylalanine decreases. The acetylated CD-immobilized membranes that have been acetylated for 2.0 h show the highest enantioselectivity of 1.33. Simple molecular modeling was also used to analyze the effects of acetylation on CD’s structure and binding efficiency to phenylalanine. It is found that acetylation would decrease the binding cability of CD but increase the chiral discrimination. Acknowledgment. We thank A-star and the National University of Singapore (NUS) for funding this research (grant no. R-279-000-164-305). LA7026384