Separation of Rebaudiana A from Steviol glycoside using a polymeric adsorbent with multi-hydrogen bonding in a non-aqueous system

Journal of Chromatography B, 971 (2014) 141–149

Contents lists available at ScienceDirect

Journal of Chromatography B journal homepage: www.elsevier.com/locate/chromb

Separation of Rebaudiana A from Steviol glycoside using a polymeric adsorbent with multi-hydrogen bonding in a non-aqueous system Jing Ba, Na Zhang, Lijuan Yao, Ning Ma, Chunhong Wang ∗ Key Laboratory of Functional Polymer Materials, Ministry of Education, Institute of Polymer Chemistry, Collaborative Innovation Center of Chemical Science and Engineering, Nankai University, Tianjin 300071, China

a r t i c l e

i n f o

Article history: Received 8 March 2014 Accepted 7 September 2014 Available online 16 September 2014 Keywords: Polymeric adsorbent Multi-hydrogen bond Synergistic effect Rebaudiana A Steviol glycosides Non-aqueous system

a b s t r a c t Rebaudioside A (RA) and stevioside (SS) are the primary effective glycoside components in Stevia Rebaudiana. The RA glycoside is sweeter, and it tastes similarly to sucrose. Because extracts with a high RA content can be used as natural sweeteners for food additives approved by the FAO and FDA, RA should generate high market demand. In this study, an efficient method for separating RA was established based on the synergistic multi-hydrogen bonding interaction between a polymeric adsorbent and the RA glycoside. To overcome the destruction of the hydrophobic affinity required for the selective adsorption of RA, an innovative non-aqueous environment was established for adsorption and separation. To this end, an initial polymeric adsorbent composed of a glycidyl methacrylate and trimethylolpropane trimethacrylate (GMA-co-TMPTMA) copolymer matrix was synthesized, and polyethylene polyamine was employed as a functional reagent designed to react with the epoxy group on GME-co-TMPTMA to form a highly selective macroporous adsorbent. The effects of the different functional reagents and the solvent polarity on the adsorption selectivity for RA and SS, respectively, were investigated. Matching the structure of the polyethylene polyamine and sugar ligand on the glycoside molecule was essential in ensuring that the maximum synergistic interaction between adsorbent and adsorbate would be achieved. Moreover, the hydrogen-bonding force was observed to increase when the polarity of the adsorption solvent decreased. Therefore, among the synthesized macroporous polymeric adsorbents, the GTN4 adsorbentbonding tetraethylenepentamine functional group provided the best separation in an n-butyl alcohol solution. Under the optimized gradient elution conditions, RA and SS can be effectively separated, and the contents of RA and SS increased from 33.5% and 51.5% in the initial crude extract to 95.4% and 78.2% after separation, respectively. Compared to conventional methods, the adsorption–desorption process is more advanced due to its procedural simplicity, low cost and adaptability for industrial production. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Steviol glycosides (SGs) are a new type of natural sweeteners extracted from Stevia Rebaudiana that contain various ent-kaurene type diterpene glycosides, including rebaudioside A (RA), stevioside (SS), rebaudioside C (RC) and dulcoside A (DA). The structure and the contents of these SGs are listed in Fig. 1 [1–4]. Among these active components, SS is the most abundant, whereas RA tastes the best; furthermore, the taste of RA is 350 times sweeter than that of sucrose but is the most similar to the taste of sucrose [5–7]. SGs have numerous applications in the food and pharmaceutical industry due to their low caloric contents, high

∗ Corresponding author. Tel.: +86 2223503935; fax: +86 2223503935. E-mail address: [email protected] (C. Wang). http://dx.doi.org/10.1016/j.jchromb.2014.09.004 1570-0232/© 2014 Elsevier B.V. All rights reserved.

sweetness, high stability [8–10] and various biological activities, such as their anti-hypertensive, anti-inflammatory, anti-tumor and immunomodulatory activities [11–15]. Consequently, in June 2008, the JECFA authorized the use of S. glycosides as food additives [16]. In December of the same year, the FDA also confirmed the safety of RA. Therefore, highly purified RA glycoside is highly popular. According to the literature [13], RA and SS glycoside are the primary bio-active components of SGs, accounting for more than 90% of the total composition of SGs. Thus, in this study, we focused our attention on the separation of RA and SS glycoside. The separation and purification of RA glycosides is hindered by the structural similarities between the RA glycoside and the other glycosides; furthermore, the acquisition of highly purified RA glycoside remains an urgent problem. To date, various methods have been used to separate the RA and SS glycosides: microwave-assisted extraction (MAE) [17], high-performance

142

J. Ba et al. / J. Chromatogr. B 971 (2014) 141–149

Fig. 1. Structure of main active components in Steviol glycosides.

liquid chromatography (HPLC) [18–20], supercritical fluid extraction [21], high-speed counter-current chromatography (HSCCC) [22], preparative chromatography [23] and the use of mixed bed resins [24,25]. However, none of these methods meet the requirements for large scale industrial production, due to their complicated procedures, low efficiency and high costs. In recent years, adsorption methods based on polymeric adsorbents have been used extensively for the extraction of natural products due to the simplicity, low-cost, high efficiency and nonpolluting nature of this technique [26–28]. However, separating RA from SGs using a typical polystyrene macroporous adsorbent is difficult because this adsorbent exhibits poor adsorption selectivity due to its hydrophobic adsorption mechanism. To improve the adsorption selectivity of the adsorbent toward RA, a special interaction must be introduced between the adsorbent and RA by changing the structure of the former. Fig. 1 shows that the number of sugar ligands in RA exceeds that in SS, strengthening the hydrogen-bonding force of RA relative to that of SS. Therefore, we can utilize the difference in hydrogen bonding by rationally designing the structure of the adsorbent and optimizing the conditions for separating the RA and SS glycosides. Chen et al. [29] introduced carbonyl groups to generate a weak hydrogen-bonding interaction between the adsorbent and SGs. Although the adsorption selectivity of the adsorbent with carbonyl groups toward RA was improved as expected, RA could not be completely separated from the SGs. Specifically, the synthesized adsorbent did not exhibit the expected adsorption selectivity toward RA, even if the hydrogen-bonding groups were introduced into the adsorbent matrix because in an aqueous solution, the hydrophobic interaction partly obscures the adsorption selectivity produced by the hydrogen-bonding interaction [30]. In addition, because water molecules are good hydrogen-bond acceptors and donors, the hydrogen-bonding interaction between the adsorbent and the SGs should be clearly weakened due to the interference of water. Therefore, in this study, we set out to resolve the two major shortcomings mentioned above, while capitalizing on the difference in hydrogen-bonding ability to separate RA and SS with high capacity and good selectivity. To improve the selectivity of the adsorbent while maintaining a high adsorption capacity, the rigid and porous structure of the adsorbent remains necessary to ensure that the adsorbent has a high specific surface area and that the functional groups are easily accessible for efficient use. Therefore, the adsorption capacity was much higher than that of the typical adsorbent containing a hydrogen-bonding group: polyamide. Some of the objectives of our study were the following: (1) use an adsorbent with a hydrophobic matrix to separate RA from SGs without destroying the adsorption selectivity through hydrophobic interactions; (2) overcome the interference of water molecules during hydrogen-bonding interaction; (3) and introduce a synergistic interaction involving multiple hydrogen bonds to increase the strength of the interaction between the RA and SS glycosides and the adsorbent. Although the amounts of sugar ligands in the RA and SS glycosides are different, the two glycosides can interact with the functional group of the adsorbent by forming hydrogen bonds; therefore, relying on a single hydrogen bond will not be enough to

separate the RA and SS glycosides. The amount of sugar ligands in the RA glycoside exceeds that of SS, and RA can form more hydrogen bonding sites, increasing the adsorption binding force with the adsorbent. Therefore, introducing a suitable functional group and ensuring its synergistic interaction remains an important problem in designing an adsorbent structure. In our adsorption experiment, non-aqueous conditions were investigated during the separation process. These conditions should weaken the effect of hydrophobic affinity on adsorption while overcoming the interference of water in the hydrogen-bonding interactions between the adsorbent and RA. To match the dissolution strength of weakly polar SGs, a series of alcohols with different polarities, such as methanol, ethanol, isopropanol and n-butyl alcohol, as well as water were employed to replace alcohol as the adsorption solution, and the effect of the polarity of each alcohol on adsorption was investigated. In addition, to obtain synergistic multiple hydrogen bonding, polyethylene polyamines were selected as the functional groups. We used glycidyl methacrylate (GMA) as a monomer and trimethylolpropane trimethacrylate (TMPTMA) as a cross-linking reagent during polymerization. Due to their similar polarities, GMA and TMPTMA can form a relatively homogeneous and highly crosslinked porous structure. Consequently, the epoxide functional group can be homogeneously distributed in the inner surface of the adsorbent. We then used polyethylene polyamines to synthesize a macroporous adsorbent with amino groups. The effects of the types and contents of the functional groups on hydrogen-bonding interactions were investigated, and the adsorption capacity and thermodynamic properties in different solvents were determined. Finally, during the gradient desorption process, each glycoside could be desorbed successively by adjusting the polarity of the solution and optimizing the desorption process. The separation of the RA and SS glycosides was remarkable, and this method was demonstrated to be suitable for large-scale production. 2. Experimental 2.1. Materials Trimethylolpropane trimethacrylate (TMPTMA, analytical reagent grade) was obtained from Anbang Chemical Ltd. (Hefei, China). Glycidyl methacrylate (GMA), 2,2 -azobisisobutyronitrile (AIBN), butyl acetate, heptane, sodium chloride, sodium hydroxide, cyclohexanol, ethanol, isopropanol, n-butanol, acetone and polyethylene polyamine were purchased from Tianjin Chemical Co. (Tianjin, China) and were of analytical grade. The crude S. glycoside extract was obtained from Kexing Trading Ltd. (Shanghai, China). The rebaudioside A (RA) and stevioside (SS) contents were 33.6% and 51.2%, respectively. The RA and SS standards were obtained from the Tianjin Institute for Drug Control. The methanol and acetonitrile used were of HPLC grade and were purchased from Concord Technology (Tianjin, China). All solutions prepared for HPLC analysis were filtered through 0.45-␮m nylon membranes before use. 2.2. Synthesis of the adsorbents 2.2.1. Synthesis of the adsorbent with epoxy group An adsorbent with an epoxy functional group was prepared using a suspension polymerization method. An organic solution composed of TMPTMA, GMA, a porogenic agent (the ratio of butyl acetate and heptane is 2:1) and AIBN (0.5%, w/w) was mixed with an aqueous solution composed of polyvinyl alcohol (PVA, 1%, w/w) and sodium chloride (5%, w/w) in a 2000-mL three-necked round-bottomed flask equipped with a mechanical stirrer, a reflux condenser and a thermometer. The volumetric ratio of the aqueous solution to the organic solution was 3:1. The round-bottomed flask was heated using a programmed heater. The mixture was stirred

J. Ba et al. / J. Chromatogr. B 971 (2014) 141–149

to generate a suspension of suitably sized oil beads in the aqueous solution. The mixture was heated to 65 ◦ C for 4 h before being held at 85 ◦ C for 6 h. The adsorbent beads were removed by filtration and washed with a large amount of hot water after washing with acetone. The proportion of TMPTMA relative to GMA was varied during the synthesis process to obtain polymeric adsorbents with different specific surface areas and epoxy contents. The adsorbents were named GT-1, GT-2, GT-3, GT-4 and GT-5, corresponding to 90%, 80%, 70%, 60% and 50% GMA contents by weight, respectively.

2.2.2. Synthesis of adsorbents with different functional groups The GT-3 adsorbent was selected as the precursor for the target GTN adsorbent with the hydrogen-bonding group in this study. The beads were air-dried and swollen with dimethylformamide (DMF) for 8 h before being mixed with 5 equivalent of functional reagent. The functional reagents were ethanediamine, diethylenetriamine, triethylenetetramine and tetraethylenepentamine, respectively. In a 500-mL three-necked round-bottomed flask equipped with a mechanical stirrer, reflux condenser and thermometer, the mixture was heated to 110 ◦ C for 6 h before being filtered; the solid was washed with deionized water and air-dried. The synthesis process used to prepare the GTN adsorbent is illustrated in Fig. 2.

2.3. Determination of the physical parameters of the adsorbents 2.3.1. Determination of pore structure The pore structure parameters of the adsorbents synthesized were measured using an automatic surface area analyzer (Autosorb-1-MP, Quantachrome Instruments, Boynton Beach, FL, USA) based on the BET nitrogen adsorption method.

2.3.2. Determination of the moisture content imbedded in the adsorbents The adsorbents hydrated with deionized water were weighed before being dried in an oven at 110 ◦ C until they reached a constant weight. The following equation was used to calculate the moisture content of the adsorbent. ˛=

Wwet − Wdry Wwet

× 100%

(1)

where ˛ is the moisture content imbedded in the adsorbent (%), Wwet the weight of the hydrated adsorbent (g) and Wdry is the weight of the dry adsorbent (g).

2.3.3. Determination of the adsorbents’ epoxy content The epoxy group content of the adsorbents was measured as previously described [31,32]: 0.2 g hydrated test adsorbent was placed in a 100-mL flask with a lid, and 25 mL saturated sodium thiosulfate solution was added. The flask was shaken for 24 h at 40 ◦ C until adsorption equilibrium was reached. Next, the 10-mL equilibrium adsorption solution was removed and titrated with an aqueous HCl solution of known concentration. The epoxy content was calculated using the following equation: E=

2.5 × CHCl × VHCl Wwet (1 − ˛)

(2)

where E is the amount of epoxy groups bound to the tested adsorbent (mmol/g), CHCl the concentration of the HCl aqueous solution (mol/L) and VHCl is the volume of the HCl aqueous solution (mL).

143

2.3.4. Determination of the amount of amino groups in the adsorbents The amount of amino groups in the synthesized adsorbent was determined using an established method and calculated according to the following equation [33]: 50CHCl − VNaOH CNaOH Wwet (1 − ˛) × 1000

N=

(3)

where N is the amount of amino groups on the tested adsorbent (mol/g), VNaOH the volume of the standard NaOH aqueous solution (mL) and CHCl and CNaOH are the concentrations of the standard HCl and NaOH aqueous solutions (mol/L), respectively. 2.4. HPLC analysis of SS and RA The weight of RA or SS in the sample was quantified by external standard method and analyzed by using HPLC [34]. The SS and RA standards were dissolved in acetonitrile at 2.0 mg/mL. Analysis was carried out using an Agilent 1200 series HPLC system equipped with a 1200-diode-array detector and a 1200 series quaternary pump with a degasser. The analysis was performed on a Kromasil NH2 column (250 mm × 4.6 mm i.d., 5 ␮m). The mobile phase, acetonitrile–H2 O (75:25, V/V), was filtered through 0.45-␮m membrane filter and ultrasonically degassed before use. The detection wavelength was 210 nm. The flow rate was 1.0 mL/min, and the injection volume was 20 ␮L. The column system was maintained at 30 ◦ C using an Agilent 1200 thermostated column compartment. The retention times of SS and RA were 7.5 and 11.0 min, respectively. 2.5. Static adsorption test To investigate the effect of the adsorption force of the adsorbent on RA and SS, a sample solution containing RA and SS was prepared, enabling the measurement of the adsorption ratio of the adsorbent relative to RA and SS. The solubility of the S. glycosides in an alcoholic solution was good compared to that in the other solvents, but hydrogen bonding is affected by water or alcohol. Therefore, to choose an appropriate solvent in which the solubility of S. glycosides is high and hydrogen bonding is a dominant interaction, we investigated the static adsorption of RA and SS glycoside with GTN-4 in solutions of water and primary alcohols. First, 0.4 g of the RA and 0.6 g of the SS were mixed and dissolved in the selected solvent and diluted to 250 mL. The concentrations of RA and SS were 1.6 and 2.4 mg/mL, respectively. The different solvents with different polarities were water, methanol, ethanol, isopropanol and n-butanol. The static adsorption tests were carried out as follows: 0.2 g of the tested adsorbent was placed into a 100-mL flask with a lid, and a 30 mL sample solution (the adsorption solution) was added. The flask was shaken (120 rpm) at 30 ◦ C until adsorption equilibrium was reached. The concentrations of RA and SS in the solution after adsorption were analyzed by HPLC. The following equations were used to quantify the equilibrium adsorption capacity, adsorption ratio and adsorption selectivity coefficient, respectively. (C0 − Ce ) × Vi Wwet × (1 − ˛)

(4)

A=

C0 − Ce × 100% C0

(5)

K=

ARA ASS

(6)

Qe =

where Qe is the equilibrium adsorption capacity of the tested adsorbent (mg/g), A the adsorption ratio (%), K the adsorption selectivity coefficient, ARA and ASS are the adsorption ratios of RA and SS,

144

J. Ba et al. / J. Chromatogr. B 971 (2014) 141–149

Fig. 2. Scheme of synthesis of GTN adsorbent: (A) adsorption in H2 O, (B) adsorption in methanol, (C) adsorption in ethanol, (D) adsorption in isopropanol and (E) adsorption in n-butanol.

respectively, C0 and Ce are the initial and equilibrium concentrations of RA or SS, respectively (mg/L), and Vi is the volume of the adsorption solution (mL). In this static adsorption experiment, the effect of the structure of the functional groups on the adsorption properties was determined before the effect of the polarity of the solvent was established. 2.6. Determination of the adsorption isotherms in different solvents The adsorption isotherms were determined as follows: 0.2 g of the hydrated tested adsorbent was placed into a 100-mL flask with a lid, and 30 mL of an aqueous RA solution of a known concentration, C0 (mg/mL) (from 1 to 5 mg/mL at 1-mg/mL intervals), was added to each flask. The flasks were shaken for 12 h at 293, 303 and 313 K until equilibrium was reached. The concentration of RA in the equilibrium solution was determined using HPLC. The equilibrium adsorption capacity Qe (mg/g) was calculated according to Eq. (4). 2.7. Dynamic adsorption and desorption tests Five grams of crude S. glycoside extract was dissolved in 1 L of n-butyl alcohol solution to prepare the adsorption solution. The concentrations of RA and SS were 1.68 and 2.56 mg/mL, respectively. Dynamic adsorption and desorption tests were carried out as follows: the hydrated test adsorbent was packed into a glass column (diameter 10 mm), and the bed volume (BV) of the adsorbent was 30 mL. The adsorption solution flowed through the adsorbent column at 1.0 BV/hr. The concentrations of RA and SS in the effluent were monitored using HPLC analysis. After adsorption, the column was washed with n-butyl alcohol and desorbed through a gradient elution method. The concentration of RA or SS in the eluent collected at 10-mL intervals was determined via HPLC. A desorption curve was obtained by plotting the concentration of SS or RA versus the volume of the desorption solution. 2.8. Product yield and repeated tests The dynamic adsorption and desorption test was conducted five times under the optimized adsorption and desorption conditions based on the above-described experiments. Two types of desorption solutions were collected under the two gradient eluent conditions and concentrated to dryness by removing the solvent under vacuum. The product corresponding to the first desorption

solution, in which the main saponin was SS, was named product I. The product corresponding to the second desorption solution, in which the main saponin was RA, was named product II. The products were analyzed by HPLC. The following equation was used to calculate the recovery: R=

m × 100% M

(7)

where R is the recovery (%), M the weight of SS or RA loaded into the selected adsorbent and m is the weight of SS or RA in the corresponding product, respectively. The standard deviation was calculated using five parallel experimental results, and the reproducibility of the separation experiments was evaluated. Furthermore, the dynamic adsorption and desorption experiments were taken under the optimized conditions respectively, using the adsorbents synthesized in different bed volume batches. 3. Results and discussion 3.1. Preparation and characterization of GT- and GTN-series adsorbents The specific surface area and the epoxy content of the adsorbents synthesized as described in Section 2.2.1 are listed in Table 1. The specific surface area of the adsorbent increased when the feed ratio of GMA was reduced during the synthesis of GT-series adsorbents. However, the change in the epoxy group content was not regular. Along with the decrease in the feed ratio of the GMA, the epoxy group content should decrease accordingly, except for that of the adsorbents GT-1 and GT-2. The epoxy contents of GT-1 and GT-2 were lower than the epoxy content of GT-3 and were even lower than the epoxy content of GT-5 because the increase in the feed ratio of GMA caused the crosslinking degree of the synthesized adsorbent to fall too low to maintain the rigid matrix structure of the adsorbent. Consequently, the collapse of the adsorbent skeleton placed some epoxy groups in a compact environment within the adsorbent matrix, hindering their reaction with organic amines in the following reaction. An adsorbent (such as GT-3) with a suitable degree of crosslinking could impart a good three-dimension porous structure, enhancing the reactivity of the epoxy groups on the porous surface. Consequently, we chose GT-3 to carry out the subsequent functionalization reaction. A GTN series of adsorbents with different amino functional groups was prepared as described in Section 2.2.2. The amino and moisture contents of the GTN adsorbents are listed in Table 2. The moisture content of the adsorbents increased with the amino group content, indicating that the hydrophobicity of the

J. Ba et al. / J. Chromatogr. B 971 (2014) 141–149

145

Table 1 Physical parameters of the synthesized GT adsorbents. Number of adsorbent

GMA content (%, w/w)

GT-1 GT-2 GT-3 GT-4 GT-5

90 80 70 60 50

Particle diameter (mm)

Epoxy group content (mmol/g)

Specific surface area (m2 /g)

0.2–0.3

0.278 0.930 1.314 1.151 0.987

117 267 386 449 488

Table 2 Physical parameters of the synthesized GTN adsorbents. Number of adsorbent

Functional group

Moisture content (%, w/w)

Content of amino group (mmol/g)

GTN1 GTN2 GTN3 GTN4

–HNCH2 CH2 NH2 –HN(CH2 CH2 NH)2 H –HN(CH2 CH2 NH)3 H –HN(CH2 CH2 NH)4 H

65.1 66.7 69.5 70.3

3.22 4.49 5.93 6.61

adsorbents decreased from GTN1 to GTN4. In addition, although the amino content per mole of amination reagent increased, the amino content of the adsorbent did not increase proportionally. For example, the mole ratio of the amino groups on diethylenetriamine and ethanediamine was 2:1, but the amount of amino groups in GTN2 was 1.39 times higher than that in GTN1. Due to steric hindrance, the amination reagent molecule could not react completely with the epoxy groups when it increased in size. However, the probability that one mole of amination reagent reacted with 2 or more moles of epoxy groups perhaps increased simultaneously. 3.2. Effect of different amino functional groups on adsorption capacity toward RA and SS The adsorption capacity of the adsorbent with different amino functional groups toward RA and SS in ethanol was investigated as described in Section 2.5, and the results are listed in Table 3. As shown in Table 3, the adsorption capacities of RA and SS both increased when increasing the amino content in the adsorbent from GTN1 to GTN4. Therefore, in an ethanol solution, the hydrophobic affinity between the adsorbent and RA or SS was clearly weakened, whereas the hydrogen-bonding interaction between the amino groups and RA or SS became the main driving force for adsorption. However, according to Table 3, the adsorption ratio (calculated according to Eq. (5)) of RA on all of the GTN adsorbents was higher than that of SS. Therefore, the adsorbent had a stronger adsorption affinity toward RA based on the hydrogen-bonding interaction because the RA molecule had more sugar ligands and could accommodate additional hydrogen-bonding interactions. Finally, the most important result was that although the content of amino groups in GTN4 was slightly higher than that in GTN3, the adsorption capacity of GTN4 toward RA was clearly twice that toward GTN3. An interaction based on multi-hydrogen bonding might have been formed when the amino group had a suitable length and structure (such as tetraethylenepentamine in the GNT4 adsorbent), resulting in a distinct increase in the adsorption affinity between GNT4 and RA. Moreover, we also observed that this multi-hydrogen-bonding interaction between GNT4 and SS was weakened relative to that between GTN4 and RA due to the different sugar ligands at R2 of RA or SS (as shown in Fig. 1). Therefore, the adsorption selectivity toward RA and SS glycosides on GTN4 varies greatly. According to Table 3, the adsorption capacity of the GNT4 adsorbent toward RA or SS was relatively low, perhaps this because the ethanol molecule can still destroy the hydrogen-bonding interaction between GNT4 and RA or SS via the polar interactions with the adsorbent or adsorbate.

3.3. Effect of different solvent on adsorption capacity of GTN4 adsorbent Because the solvent polarity could affect the interaction of the GTN4 adsorbent, the adsorption capacity of GTN4 toward RA and SS in different solvents was monitored as described in Section 2.5, and the results are listed in Table 4. Table 4 indicates that the adsorption capacity of RA clearly increased with the decrease in solvent polarity. When water was used as the solvent, it remarkably destroyed the hydrogen bonding between amino groups on GTN4 and RA because it acted as a good hydrogen bonding donator and acceptor; therefore, the adsorption capacity of the adsorbent was lowest in an aqueous solution. When decreasing the solvent polarity, the interference with the hydrogen-bonding interaction decreased, and the adsorption capacity of RA increased appreciably. Table 4 indicates that although the adsorption capacities of SS and RA glycoside increased with the decrease insolvent polarity, the adsorption selectivity coefficient K (calculated according to Eq. (6)) of RA to SS increased markedly because in a highly polar solvent, hydrophobic interactions should become the key driving force for the adsorption of GTN4. Because RA exhibited the same hydrophobic structure as SS, the adsorption capacity of GNT4 toward RA was similar to that of SS even though GNT4 could produce a hydrogenbonding interaction with RA and SS. Specifically, that hydrophobic interaction obscured the differences in the adsorption selectivity of GNT4 toward RA and SS. When the solvent polarity decreased, the hydrophobic interaction between GNT4 and RA or SS decreased, and the selectivity of GNT4 toward RA and SS adsorption based on the hydrogen-bonding interactions began to appear. Therefore, the adsorption selectivity coefficient for RA to SS increased appreciably with the decrease in the polarity of the adsorption solvent. Consequently, in weakly polar n-butanol, GNT4 exhibited not only the highest adsorption capacity toward RA but also the highest selectivity coefficient of RA to SS. This exciting result demonstrates that the multi-hydrogen-bonding synergistic effect in a non-aqueous system is reasonable and feasible. 3.4. Studies of adsorption thermodynamics of GNT4 adsorbent toward RA The thermodynamics describing the adsorption of RA onto the GTN4 adsorbent were investigated as described in Section 2.6, and the corresponding adsorption isotherms are illustrated in Fig. 3. The isosteric adsorption enthalpies were calculated based on the Van’t Hoff equation [28,35]: ln Ce =

Hm + Ka RT

(8)

Qe(mg/g)

J. Ba et al. / J. Chromatogr. B 971 (2014) 141–149

Qe(m g/g)

146

15

50

40

10

30

40 ºC 30 ºC 20 ºC

5

ºC ºC ºC

20

10 1

2

3

4

1

5

2

3

4

Ce( mg/m L)

(A) Adsorption in H2 O

(B) Adsorption in methanol

Qe(mg/g)

500

400

400 350 300

300

250

ºC ºC ºC

200

ºC ºC ºC

200 150

100 1

2

3

0.5

4

1.0

1.5

2.0

Ce(mg/mL)

1000

800

600

ºC ºC ºC

400

0.5

3.0

3.5

4.0

4.5

(D) Adsorption in isopropanol

1200

200 0.0

2.5

Ce(mg/mL)

(C) Adsorption in ethanol Qe(mg/g)

Qe(mg/g)

5

Ce(mg/mL)

1.0

1.5

2.0

2.5

3.0

3.5

Ce(mg/mL)

(E) Adsorption in n-butanol Fig. 3. Adsorption isotherms of GTN4 towards RA in different solvent.

J. Ba et al. / J. Chromatogr. B 971 (2014) 141–149

147

Table 3 Adsorption capacity of GTN series adsorbent towards RA and SS. Adsorbent number

C0 (mg/mL)

GTN1 GTN2 GTN3 GTN4

Ce (mg/mL)

Qe (mg/g)

A (%)

RA

SS

RA

SS

RA

SS

RA

SS

1.60

2.40

1.46 1.41 1.38 1.18

2.22 2.18 2.14 2.11

60.2 85.6 108.1 212.1

77.4 99.1 127.9 146.5

8.7 11.8 13.7 26.2

7.5 9.2 10.8 12.1

Table 4 Adsorption capacity of GNT4 towards RA and SS in different solvents. Solvent

H2 O Methanol Ethanol Isopropanol n-Butanol

C0 (mg/mL)

Ce (mg/mL)

Qe (mg/g)

A (%)

K

RA

SS

RA

SS

RA

SS

RA

SS

1.60

2.40

1.59 1.55 1.18 0.949 0.554

2.39 2.34 2.11 1.96 1.90

3.64 25.2 212.1 328.8 528.3

5.05 30.3 146.5 222.2 262.6

0.450 3.12 26.2 40.7 65.4

0.417 2.50 12.1 18.3 20.7

where Hm is the isosteric adsorption enthalpy (kJ/mol), Ce the equilibrium concentration of RA, T the adsorption temperature (K) and R and Ka are constants. Hm was calculated from the slope of the line plotting ln Ce versus 1/T; the calculated isosteric enthalpies of RA adsorption onto the GNT4 adsorbent in different solvents are listed in Table 5. Table 5 reveals that the value of the isosteric adsorption enthalpy increased with a decrease in solvent polarity. In addition, Hm was positive in an aqueous solution, indicating that the adsorption was endothermic. The adsorption enthalpy was negative in the other solutions, demonstrating that the adsorption was exothermic. Generally, if RA adsorbed onto GNT4, the interaction between GNT4 and the solvent molecules must be destroyed. When the solvent was H2 O and could easily form hydrogen-bonding interactions with GNT4, additional energy had to be provided to destroy the interaction between H2 O and GNT4; the amount of energy exceeded that released when GNT4 interacts with RA. Therefore, Hm was positive in aqueous solutions, gradually becoming negative with a decrease in solvent polarity. Consequently, the adsorption capacities of GNT4 toward RA highlighted the differences between the solvents, as listed in Table 4. n-Butanol was selected to carry out the adsorption of RA onto GNT4. In addition, because the adsorption enthalpy in the aqueous solution was positive, H2 O could be used as a desorption reagent to elute RA and SS after adsorption. With a decrease insolvent polarity, the ability of the solvent to destroy the hydrogen-bonding interactions with GNT4 was weakened. Therefore, it was determined that we could provide a suitable gradient to elute RA and SS by adjusting the ratio of the solvents with different polarities. This deduction was proven correct in the gradient elution test the followed.

concentration of RA or SS. The results are illustrated in Fig. 4. The leakage point was defined as the 1% ratio of the exit to the inlet SS concentration. As shown in Fig. 4, the processing volume of the adsorption solution prepared as described in Section 2.7 for the GNT4 adsorbent was 210 mL (7 BV). After the dynamic adsorption, gradient desorption of the RA and SS loaded onto GNT4 was carried out. To achieve the best separation of RA and SS, two types of gradient conditions were tested, and the corresponding gradient desorption curves are illustrated in Fig. 5. As shown in Fig. 5, the adsorption affinity toward the SS glycoside was weaker than that toward the RA glycoside. After elution with a 5% H2 O–95% ethanol solution, almost all of the SS glycoside loaded on the GNT4 was desorbed by 140 mL of the 5% H2 O–95% ethanol solution; we also observed that a small portion of the RA glycoside was desorbed with SS volumes of 5% H2 O–95% ethanol of up to 110 mL (as illustrated in Fig. 5A). Therefore, the desorption ability of water was slightly strong, allowing for the desorption of RA with SS. In turn, a known concentration of methanol was selected as the desorption solution; the elution curve of 5% methanol–95% ethanol solution indicates that SS can be desorbed completely with 150 mL of 5% methanol–95% ethanol solution, whereas RA can be retained on the adsorbent column. When the methanol concentration increased, RA could be desorbed easily by a stronger solution, such as 10% methanol–90% ethanol solution, as illustrated in Fig. 5B.

3.5. Dynamic adsorption and desorption Dynamic adsorption was carried out as described in Section 2.7. The leakage curve for the GNT4 adsorbent toward RA or SS was obtained by plotting the volume of the effluent solution and the

Table 5 Isosteric adsorption enthalpy of GTN4 towards RA in different solvent. Solvent

Hm (kJ/mol)

Ka

R2

H2 O Methanol Ethanol Isopropanol n-Butanol

15.2 −11.6 −18.8 −21.5 −36.3

−4.92 5.47 8.16 9.34 15.1

0.9974 0.9943 0.9989 0.9937 0.9984

1.08 1.25 2.16 2.22 3.16

Fig. 4. Leakage curves of GNT4 towards RA and SS.

148

J. Ba et al. / J. Chromatogr. B 971 (2014) 141–149

Fig. 5. Gradient desorption curves of GNT4 towards RA and SS under different conditions (A): eluent was 5% H2 O–95% C2 H5 OH 150 mL, and then 10% H2 O–90% C2 H5 OH 150 mL. (B) Eluent was 5% CH3 OH–95% C2 H5 OH 150 mL, and then 10% CH3 OH–90%C2 H5 OH 150 mL.

Table 6 Results of the separation of RA and SS on GNT4 adsorbent. Test no.

Initial sample a

1 2 3 4 5 Average SDe

Product I b

c

W (mg)

Product II

WRA (mg)

WSS (mg)

352.8

537.6

640.8 642.5 637.4 641.1 638.2

WRA (mg) 45.8 47.6 45.2 44.4 46.1

503.7 500.5 497.8 502.6 497.8

WSS (mg)

– –

– –

– –

– –

– –

RSS (%)

Purity of SS (%)

W (mg)

WRA (mg)

WSS (mg)

RRA (%)

Purity of RA (%)

93.7 93.1 92.6 93.5 92.6

78.6 77.9 78.1 78.4 78.0

287.8 292.4 288.0 288.9 288.7

275.4 276.6 274.5 276.2 276.6

16.1 15.6 17.2 16.9 14.8

78.1 78.4 77.8 78.3 78.4

95.7 94.6 95.3 95.6 95.8

93.1 0.005

78.2 0.003

– –

– –

– –

78.2 0.003

95.4 0.003

a

WRA is the weight of RA in the sample. WSS is the weight of SS in the sample. c W is the weight of the product after separation. d RSS or RRA is the recovery of SS or RA corresponding to the product I or II, respectively. e SD is standard deviation. b

3.6. Results of the separation of RA and SS glycoside on the GTN4 adsorbent Our results reveal the optimization of the separation process of RA and SS on GTN4 as follows: 3.6.1. Adsorption The concentrations of RA and SS in the adsorption solution were 1.68 and 2.56 mg/mL, respectively; the adsorption solvent was the n-butyl alcohol solution; the process volume was 7 BV (1 BV = bed volume of adsorbent in adsorption column); the adsorption rate was 1 BV/h; adsorption was carried out at room temperature. 3.6.2. Desorption The desorption of 5 BVs of the 5% methanol–95% ethanol solution was carried out as the first desorption step, followed by that of 4 BVs of the 10% methanol–90% ethanol solution as the second desorption step; the desorption rate was 0.5 BV/h; desorption was carried out at room temperature. The adsorption and gradient desorption test was carried out under the above-described conditions. The product was obtained, and repeated tests were completed as described in Section 2.8. The product obtained during the first step of the gradient desorption was named product I, whereas that obtained from the second step was named product II. The initial crude extract of the S. glycosides

(products I and II) was determined using HPLC, as described in Section 2.4, and the results are listed in Table 6. As shown in Table 6, after treatment with GNT4, the RA content increased from 33.5% in the initial crude extract to 95.4% in product II; that of SS increased from 51.5% to 78.2% in product I. The performance of GNT4 in separating RA and SS was repeatable and stable. It is known that the ratio of RA and SS may be very different in different crude extract of the S. glycosides. The ability to isolate RA and SS from different extracts is very important in industrial production. So in our cooperative factory, the separation of various extracts containing different ratio of RA and SS based on GNT4 adsorbent were carried out in large-scale commercial process (the bed volume of the GNT4 adsorbent was 1500 L). And the results are listed in Table 7. For all of the selected crude extract of the S. glycosides, GNT4 adsorbent exhibited no difference in isolation effect on RA and SS. This indicates that the purification method based on GNT4 adsorbent has reproducibility and can be applied in large scale industrial production. The chromatograms of the initial sample solution prepared as described in Section 2.5, as well as those of products I and II, are shown in Fig. 6. The chromatogram of product II shows that the highly purified RA glycoside was separated with a minimal amount of SS glycoside. Similarly, for product I, little RA glycoside was employed. To ensure that the SS glycoside was completely

J. Ba et al. / J. Chromatogr. B 971 (2014) 141–149

149

Table 7 Separation results of GNT4 for different raw material with different ratio of RA and SS Note . Test no.

1 2 3 4 5 Average SD

Initial sample

Product I

WRA (kg)

WSS (kg)

W (kg)

WRA (kg)

WSS (kg)

RSS (%)

Purity of SS (%)

W (kg)

WRA (kg)

WSS (kg)

RRA (%)

Purity of RA (%)

17.64 25.30 21.96 19.72 16.17

26.88 17.51 24.31 25.43 31.64

31.12 20.72 26.66 28.45 35.45

1.92 2.02 2.41 2.37 1.50

24.24 15.85 21.49 22.25 28.82

90.2 90.5 88.4 87.5 91.1

77.9 76.5 80.6 78.2 81.3

14.03 20.75 17.47 15.68 12.95

13.48 19.78 16.93 14.85 12.32

1.09

76.4 78.2 77.1 75.3 76.2

96.1 95.3 96.9 94.7 95.1

– –

– –

– –

– –

89.5 0.015

78.9 0.020

– –

– –

76.6 0.011

95.6 0.0088

– –

Product II

Note: The data in this table was the result of the commercial scale separation process and the bed volume of the adsorbent was 1500 L. The meaning of WRA , WSS , W, RSS , RRA and SD is same as that in Table 6.

superior due to its procedural simplicity, low cost and adaptability for industrial applications. Acknowledgements We thank the financial support by the National Natural Science Foundation of China (NSFC) for (Grant No. 21074060), Tianjin Municipal Science and Technology Commission (Grant Nos. 13JCZDJC32900, 09JCYBJC13600) and PCSIRT (IRT1257). References

Fig. 6.

desorbed, a portion of RA glycoside was unavoidably lost. During the subsequent experiments, we regulated the process of gradient desorption properly, and highly purified RA glycoside was obtained. 4. Conclusions In this study, a new method using a macroporous polymeric adsorbent was established to separate the RA and SS glycosides in a S. glycoside extract. Macroporous polymeric adsorbents with epoxide groups were synthesized, and different amino functional groups were introduced on the adsorbent. The effects of the functional groups and solvent used on the adsorption capacity toward the RA and SS glycosides were studied. The adsorption binding ability of the RA glycoside toward the adsorbent exceeded that of the SS glycoside when a synergistic multi-hydrogen-bonding interaction was utilized in a non-aqueous system. The GTN4 adsorbent exhibit the greatest ability to separate the RA and SS glycosides. Analysis of the adsorption thermodynamics showed that the absolute value of the isosteric adsorption enthalpy increased with a decrease in solvent polarity, indicating that the hydrogen bonding interaction of the adsorbents and RA glycoside occurred more easily. Through treatment under optimal conditions on a column packed with GTN4, the separation of the two glycosides was successfully achieved. Under the optimized separation conditions, the RA and SS contents increased from 33.5% and 51.5% in the initial crude extract to 95.4% and 78.2% in the corresponding products, respectively. The recoveries of RA and SS were 78.2% and 93.1%. Compared to the conventional methods used, this adsorption–desorption process is

[1] O. Tanaka, Trends Anal. Chem. 1 (1982) 246. [2] N.I. Bondarev, M.A. Sukhanova, O.V. Reshetnyak, A.M. Nosov, Biologiaplantarum 47 (2003) 261. [3] U. Wölwer-Rieck, J. Agric. Food Chem. 60 (2012) 886. [4] C. Gardana, M. Scaglianti, P. Simonetti, J. Chromatogr. A 1217 (2010) 1463. [5] M.C. Carakostas, L.L. Curry, A.C. Boileau, D.J. Brusick, Food Chem. Toxicol. 46 (2008) S1. [6] J.E. Brandle, A.N. Starratt, M. Gijzen, Can. J. Plant Sci. 78 (1998) 527. [7] S.K. Goyal, R.K. Samsher, Goyal, Int. J. Food Sci.Nutr. 61 (2010) 1. [8] G. Kroyer, J. Verbr. Lebensm. 5 (2010) 225. [9] U. Wölwer-Rieck, W. Tomberg, A. Wawrzun, J. Agric. Food Chem. 58 (2010) 12216. [10] J.F. Clos, G.E. Dubois, I. Prakash, J. Agric. Food Chem. 56 (2008) 8507. [11] M.H. Hsieh, P. Chan, Y.M. Sue, J.C. Liu, T.H. Liang, T.Y. Huang, B. Tomlinson, M.S.S. Chow, P.F. Kao, Y.J. Chen, Clin. Therap. 25 (2003) 2797. [12] V. Chatsudthipong, C. Muanprasat, Pharmacol. Ther. 121 (2009) 41. [13] S. Madan, S. Ahmad, G.N. Singh, K. Kohli, Y. Kumar, R. Singh, M. Garg, Ind. J. Nat. Prod. Resour. 1 (2010) 267. [14] K. Yasukawa, S. Kitanaka, S. Seo, Biol. Pharm. Bull. 25 (2002) 1488. [15] I. Sehar, A. Kaul, S. Bani, H.C. Pal, A.K. Saxena, Chemico-Biol. Interact. 173 (2008) 115. [16] JECFA, Joint FAO/WHO Expert Committee on Food Additives Monographs 5, in: Compendium of Food Additive Specifications from 69th JEFCA Meeting, Rome, June 17–26, 2008. [17] V. Jaitak, Bandna, B. Singh, V.K. Kaul, Phytochem. Anal. 20 (2009) 240. [18] B. Nikolova-Damyanova, V. Bankova, S. Popov, Phytochem. Anal. 5 (1994) 81. [19] U. Woelwer-Rieck, C. Lankes, A. Wawrzun, M. Wüst, Eur. Food Res. Technol. 231 (2010) 581. [20] D. Bergs, B. Burghoff, M. Joehnck, G. Martin, G. Schembecker, J. Verbr. Lebensm. 7 (2012) 147. [21] A. Erkucuk, I.H. Akgun, O. Yesil-Celiktas, J. Supercrit. Fluids 51 (2009) 29. [22] X.Y. Huang, J.F. Fu, D.L. Di, Sep. Purif. Technol. 71 (2010) 220. [23] D. Bergs, J. Merz, A. Delp, M. Joehnck, G. Martin, G. Schembecker, J. Verbr. Lebensm. 7 (2012) 295. [24] Y.F. Liu, D.L. Di, Q.Q. Bai, J.T. Li, Z.B. Chen, S. Lou, H.L. Ye, J. Agric. Food Chem. 59 (2011) 9629. [25] Z.B. Chen, X.L. Wei, J. Li, D.L. Di, Sep. Purif. Technol. 89 (2012) 22. [26] G.P. Pi, P. Ren, J.M. Yu, R.F. Shi, Z. Yuan, C.H. Wang, J. Chromatogr. A 1192 (2008) 17. [27] P. Wang, X.T. Geng, G.P. Pi, R.F. Shi, Z. Yuan, C.H. Wang, J. Chromatogr. B 878 (2010) 3375. [28] N. Ma, P. Wang, X. Kong, R.F. Shi, Z. Yuan, C.H. Wang, J. Sep. Sci. 35 (2012) 36. [29] T.H. Chen, Y. Zhang, X.H. Liu, Z.Q. Shi, J.T. Sun, B.L. He, Sci. China Ser. B: Chem. 41 (1998) 436. [30] N. Maity, G.F. Payne, Langmuir 7 (1991) 1247. [31] R. Axen, H. Drevin, J. Carlsson, Acta Chem. Scand. B 29 (1975) 471. [32] L. Sundberg, J. Porath, J. Chromatogr. A, 90 (1974) 87. [33] B.L. He, W.Q. Huang, Ion Exchanger and Adsorptive Resins, Shanghai Sci. Tech. Press, Shanghai, 1995. [34] Chinese Standard GB 8270-1999. [35] G.F. Payne, N. Maity, Ind. Eng. Chem. Res. 31 (1992) 2024.