Journal of Colloid and Interface Science 396 (2013) 165–172

Contents lists available at SciVerse ScienceDirect

Journal of Colloid and Interface Science

www.elsevier .com/locate / jc is

Multiscale study on the interaction mechanism between ginsenosidebiosurfactant and saikosaponin a

Xingxing Dai a,1, Xinyuan Shi b,c,d,⇑,1, Qianqian Yin a, Haiou Ding a, Yanjiang Qiao b,c,d,⇑a School of Traditional Chinese Medicine, Capital Medical University, Beijing 100069, Chinab Beijing University of Chinese Medicine, Beijing 100102, Chinac Key Laboratory of TCM-information Engineer of State Administration of TCM, Beijing 100102, Chinad Beijing Key Laboratory for Basic and Development Research on Chinese Medicine, Beijing 100102, China

a r t i c l e i n f o a b s t r a c t

Article history:Received 19 November 2012Accepted 7 January 2013Available online 25 January 2013

Keywords:BiosurfactantGinsenosideSaikosaponin aMultiscale methodSimulation

0021-9797/$ - see front matter � 2013 Elsevier Inc. Ahttp://dx.doi.org/10.1016/j.jcis.2013.01.017

⇑ Corresponding authors at: Beijing University o100102, China. Fax: +86 10 84738661.

E-mail addresses: shixinyuan01@163.com (X. Shi),1 These authors contributed equally to this work.

Ginsenoside is an important class of saponin biosurfactant that is derived from ginseng. The interactionsbetween ginsenoside Ro, Rb1, and Rg1 with saikosaponin a (SSa) were explored using multiscale methods.The order of interaction strength was found to be Ro > Rb1 > Rg1. Ro markedly increased the solubility ofSSa; however, Rb1 could only disperse SSa solid in aqueous medium. No significant interaction wasobserved between Rg1 and SSa. Ro formed vesicles in aqueous medium while Rb1 and Rg1 formed spher-ical micelles. The differences in the available surface area of the aggregates appear to have some influenceon the interactions between ginsenoside and SSa. However, more important effects are related to theirchemical structures and interaction energy. According to the molecular simulation results, glucuronicacid linked to Ro molecules significantly reduced the potential energy through its strong electrical attrac-tion to SSa, which contributed greatly to the strong compatibility between them. The greater number ofsugars in Rb1, as compared to Rg1, created more binding sites with SSa, thus resulting in stronger inter-action between Rb1 with SSa than between Rg1 and SSa. Spherical and worm-like micelles were found tobe formed by Rb1 and SSa molecules. This was different from Ro and SSa, which formed vesicles. The for-mation of worm-like micelles was through the fusion and modification of small spherical micelles. Theseresults may guide in expanding the applications of ginsenoside.

� 2013 Elsevier Inc. All rights reserved.

1. Introduction adjuvants for delivery of agromaterials through plant cuticle mem-

Saponins are biosurfactants that are primarily found in plants.They are amphipathic molecules containing a hydrophobic triter-pene or steroid aglycone, and one or more hydrophilic sugarchains. The differences in aglycones, and the number of sugarsand sugar chains attached to the glycosidic bonds, give them di-verse surface properties, such as foaming, emulsifying, and solubi-lizing properties [1]. Coupled with a wide range of biologicalactivities, such as anticancer, anti-inflammatory and anti-choles-terol effects [2], saponins have many successful commercial appli-cations in the food, cosmetics, agriculture, and pharmaceuticalindustry. For example, quillaja saponin has been used as foamingagent (type 1 quillaja extract) and emulsifier (type 2 quillaja ex-tract) for preparations of drinks [3]. Some saponins have been usedas immunological adjuvants due to their immune enhancing prop-erties and surface activities [4–7]. Extracts of Quillaja saponariaand Balanites aegyptiaca were proposed as potential natural

ll rights reserved.

f Chinese Medicine, Beijing

yjqiao@263.net (Y. Qiao).

branes [8]. Abundant studies have been done on the solubilizationeffect of saponin. They have been used as adjuvants to enhance thesolubility or absorption of pharmacologically active substances ordrugs in pharmaceutical preparations [9–17]. Increasing consumerdemand for natural products has led to more attention on sapo-nins, the ‘‘green natural additives.’’ Studies on the interaction ofsaponins with different compounds are likely to serve as guidancefor more applications of saponins.

Ginseng (roots of Panax gingseng C. A. Mey) is an importantsaponin-containing resource that has been widely used. At least40 different kinds of ginsenosides have been isolated from ginseng[18–22]. They are classified as dammarane, oleanolic acid, andother type of saponins (their chemical structures are shown inFig. S1 in supporting information). Saponins of damarane can alsobe divided into 20(s)-protopanaxadiol saponin (such as ginseno-side Rb1) and 20(s)-protopanaxatiol saponin (such as ginsenosideRg1). The typical saponin of oleanolic acid is chikusetsusaponin V(ginsenoside Ro). It has been known that ginsenoside has a solubi-lization property which is not typical of all kinds of ginsenosides.Studies of Osamu Tanaka [12] have shown that the water solubilityof saikosaponin a (SSa), the main active component ofRadix Bupleuri (roots of Bupleurum chinense DC. and Bupleurum

166 X. Dai et al. / Journal of Colloid and Interface Science 396 (2013) 165–172

scorzonerifolium Wild.), which is co-prescribed with ginseng in thetraditional Chinese medicinal (TCM) formula, was significantlyhigher in the presence of Ro but was not influenced by the presenceof dammarane saponins. It is very interesting to know the exactunderlying mechanism that causes the difference in interactionsbetween Ro and dammarane saponins with SSa. Our previous stud-ies have explored the interaction between Ro and SSa to illustratethe solubilization mechanism of Ro [23]. In this study, ginsenosideRb1 and Rg1 are taken as models to study the interactions betweendammarane saponin with SSa. The results were compared with Roto provide a detailed understanding of the interaction mechanismof ginsenoside with other compounds. The results from this studywill potentially help to expand the applications of ginsenoside inthe pharmaceutical industry and other areas.

In this study, common experimental methods such as high per-formance liquid chromatography (HPLC), transmission electronmicroscopy (TEM), and dynamic light scattering (DLS) measure-ments were employed to give a macroscopic understanding ofthe interactions between ginsenoside and SSa. However, the inter-action mechanisms and the exact internal structures of ginseno-side and SSa aggregates are difficult to observe by thesemethods. Computerized calculation and simulation methods makethat possible, providing useful tools to extrapolate the experimen-tal results from atomic- and mesoscale levels. Compatibility be-tween different compounds plays an important role indetermining their interaction abilities. A molecular simulationmethod which combines a modified Flory–Huggins model andmolecular simulation techniques (conducted on Blends moduleprovided in Materials studio 4.1, Accelrys Inc.) was used to calcu-late the compatibility of binary mixtures (referred to in Blends,the datasheet of Materials Studio, http://accelry.com/materials-studio). The information that can be obtained includes tempera-ture-dependent interaction parameter v, thermodynamic mixingvariables (energy and free energy of mixing), binding energy, andmore. It is of great advantage in studying the compatibility of com-pounds at atomic-scale. Dissipative particle dynamics (DPD) is animportant tool to study the interaction of macromolecules atmesoscale level, especially for the drug loading systems [24–33].It is a particle-based method which coarse-grains a group of famil-iar atoms to a bead. In this way, it is able to gain larger length andtime scale relative to traditional atomistic scale [34]. It can also di-rectly represent the movement of mesomolecules. By combiningthe two simulation methods, as well as experimental methods, itis able to illustrate the interaction mechanism of ginsenoside withother compounds more comprehensively at a multiscale level.

2. Experimental

2.1. Materials

Ro, Rb1, Rg1, and SSa were purchased from National Institutesfor Food and Drug Control and Vicks biotechnology Co., Ltd. (98%purity). Saturated solutions of SSa and SSa, in equimolar concentra-tion of Ro, Rb1 and Rg1 (1.0 mmol/L), were prepared by dissolvingexcess SSa in water and Ro, Rb1, and Rg1 solution, respectively, at25 �C for 24 h. Samples for HPLC were then filtered through a0.45 lm millipore filter to remove dust and other large particles.However, in order to study the mesoscopic behavior of ginseno-side, samples of TEM and DLS were filtered through a 0.8 lm mil-lipore filter (mesoscale is 10–1000 nm).

2.2. Determination of critical micelle concentration (CMC) by surfacetension method

Surface tensions of Ro, Rb1 and Rg1 solutions, at different con-centrations, were measured with Powereach tensiometer (JK99B,

Shanghai Zhongchen Digital Technique & Equipment Co., Ltd., Chi-na), using the Wilhelmy plate method at 25 �C. The platinum platewas cleaned and heated to red/orange color before each measure-ment. A plot of surface tension versus concentration of surfactantwas produced. The surface tension first decreased rapidly and thendropped to a minimum as concentration was increased. The CMCwas determined from the inflection point.

2.3. Determination of solubility of SSa by HPLC analysis

The solubility of SSa in different ginsenoside solutions was ana-lyzed by HPLC assay recommended by Chinese Pharmacopoeia(ChP, 2010 Edition). Agilent 1100 series equipped with quaternarypump, autosampler, using a DAD detector. Analyses were con-ducted on an ODS C18 column (150 mm � 4.6 mm, 5 lm, Waters)in acetonitrile/water mobile phase, with the flow rate of 1.0 mL/min. The first gradient elution was from 25:75 (v/v) to 90:10(v/v) within 0–50 min, followed by the second gradient elutionback to 25:75 (v/v) in 10 min. The column temperature was25 �C, and the detection wavelength was set at 210 nm.

2.4. Transmission electron microscopy (TEM)

The aggregate morphologies of equimolar Ro, Rb1, and Rg1 solu-tion (1.0 mmol/L) and their mixtures with SSa were observed on atransmission electron microscope (TEM, JEOL JEM-1230 micro-scope) operating at an acceleration voltage of 80 kV. The solutionsamples were deposited on copper grids that had been precoatedwith a thin film of formvar and then coated with a thin carbon film.The liquid was blotted off with filter paper after a few minutes, andthe grids were air-dried.

2.5. Dynamic light scattering (DLS) measurement

DLS was used to monitor the interaction process of Rb1 and Rg1

with SSa in aqueous medium. Equal amount of the excess SSa wasadded to equimolar Rb1 or Rg1 solutions. The solutions were sha-ken for a period of time at 25 �C and then filtered to remove undis-solved SSa. The size changes of the aggregates in solutions, withtime elapse, were measured with a Zetasizer Nano-ZS dynamiclight scattering (DLS) instrument (Malvern, UK) at an angle of173� at 25 �C.

3. Simulation

3.1. Molecular simulation

The molecular simulations were conducted in Blends moduleusing commercial software package Materials Studio 4.1 (AccelrysInc.). This module combines a modified Flory–Huggins model andmolecular simulation techniques to calculate the compatibility ofbinary mixtures. Flory–Huggins model is calculated by the follow-ing equation:

DGRT¼ /b

nbln /b þ

/s

nsln /s þ v/b/s ð1Þ

where DG is the free energy of mixing (per mole), /i is the volumefraction of component i, ni is the degree of polymerization of com-ponent i, and v is the interaction parameter which is defined by thefollowing equation:

v ¼ Emix

RTð2Þ

where Emix is the mixing energy, that is, the difference in free en-ergy due to interaction between the mixed and the pure state. A

Fig. 1. (a) Solubility of SSa in different solutions and (b) photos of ginsenoside/SSainteraction systems in aqueous medium.

X. Dai et al. / Journal of Colloid and Interface Science 396 (2013) 165–172 167

lower value of v indicates a better compatibility between the com-ponents in the binary mixture system.

Blends module distinguishes the components by using the roleproperty. In each mixture considered by Blends, one componenthas a base role (abbreviated as b) and the other has a screen role(abbreviated as s).

3.2. DPD simulation

Dissipative particle dynamics (DPD) is a mesoscopic simulationtechnique suitable to the study of the collective behavior of com-plex fluids. A DPD bead represents a small region of fluid matterand its motion is assumed to be governed by Newton’s laws fol-lowing equation [35]:

dri

dt¼ vi ð3Þ

midvi

dt¼ f i ð4Þ

where ri, vi, mi, and fi denote the position vector, velocity, mass, andtotal force acting on particle i, respectively.The force fi betweeneach pair of beads contains three parts: a harmonic conservative

interaction force FCij

� �, a dissipative force FD

ij

� �, and a random force

FRij

� �. The expression is given as f i ¼

Pj–i FC

ij þ FDij þ FR

ij

� �. All forces

are short-range with a fixed cutoff radius rc, which is usually chosenas the reduced unit of length rc � 1. All the simulations were con-ducted in Materials studios 4.1 (Accelrys Inc.), and the simulationdetails are shown in supporting information.

4. Results and discussion

4.1. Interactions between different ginsenosides and SSa

The solubility of SSa, in equimolar concentration of Ro, Rb1 andRg1 solutions, as well as in aqueous solutions was determined byHPLC, and the results are shown in Fig. 1a. It was found that thesolubility of SSa was markedly increased by Ro, but was not signif-icantly affected by Rb1 and Rg1. These results were consistent withthose reported before [12]. The solubilization of SSa by Ro was theresult of strong interaction between Ro and SSa molecules to formsmall aggregates. However, the performances of the mixed solu-tions revealed that Rb1 could disperse the SSa solid in aqueousmedium (Fig. 1b). The interaction between Rb1 and SSa was weakerthan between Ro and SSa because a suspension is a thermodynam-ically unstable system. The removal of the large Rb1-SSa aggregatesby filtration resulted in little change in the solubility of SSa. Thephenomena in Rg1/SSa flask indicated that the interaction betweenRg1 and SSa may be similar to that between Rb1 and SSa; however,it was much weaker, and most of the SSa solid added remained un-changed at the bottom of flask. The differences in the interactionsbetween different types of ginsenoside and SSa can help lead fur-ther research of ginsenosides toward specific applications. There-fore, this work focused on the mechanism of interaction in orderto illustrate their differences.

4.2. Self-assembly behavior of ginsenoside

4.2.1. Determination of critical micelle concentration (CMC)CMC is the concentration at which the surfactant starts to self-

assemble into micelles. It reflects the micellization ability of thesurfactant [36]. The CMC values of Ro, Rb1, and Rg1 were deter-mined to be 0.34 mmol/L, 0.25 mmol/L, and 0.08 mmol/L, respec-tively (the detailed results are shown in supporting information).This result was mainly attributed to differences in the types and

number of hydrophilic sugars. Rg1 contained the fewest sugars,which resulted in the strongest hydrophobic interaction. It waseasiest for Rg1 to undergo micellization. However, Ro containedcharged glucuronic acid group. This can increase water solubilityof Ro because of its strong hydration and can also work againstthe aggregation process because of its charge repulsion [37]. As aresult, the CMC value for Ro was the largest. It was found thatthe order of CMC values of the ginsenoside was not consistent withtheir interaction strength, as indicated in Section 4.1. This suggeststhat the micellization ability of the ginsenoside was not the keyfactor affecting its interaction with SSa.

4.2.2. Aggregation morphologies of ginsenoside and SSaThe amphipathic structures of saponin molecules cause them to

self-assemble in aqueous solution to form aggregates. Their aggre-gation morphologies may have a significant impact on their inter-actions with other compounds [37]. Our previous studies showedthat Ro could form vesicles with diameters of 30–50 nm in aqueoussolution (Fig. 2a) [23]. The repulsive interaction caused by thecharged glucuronic acid located on the vesicle surface made thevesicles disperse uniformly and stably in the aqueous solution.However, Rb1 and Rg1 molecules self-assembled in aqueous solu-tion to form spherical micelles, with diameters of 50–80 nm in caseof Rb1 and 20–60 nm in case of Rg1 (Fig. 2b and c). Both Rb1 and Rg1

micelles collided with each other to form large fractal aggregatesbecause of their lower hydrophilicity as compared to Ro. That re-sulted in fewer chances for SSa molecule to interact with, andget into the Rb1 and Rg1 micelle, especially the micelles inner tothe aggregates. However, there was a large superficial area of Rovesicles that was available for SSa molecules to penetrate. This pos-sibly contributed to the differences in interaction effects.

4.3. Study on ginsenoside and SSa blending systems by molecularsimulation

The morphology results could partially explain the interactiondifference between ginsenoside and SSa, but still could not explainthe difference entirely. Besides that, some other questions about

Fig. 2. TEM images of 1.0 mmol/L ginsenoside solution and saturated solution of SSa: (a) Ro vesicles; (b) Rb1 micelles; (c) Rg1 micelles; and (d) SSa micelles.

168 X. Dai et al. / Journal of Colloid and Interface Science 396 (2013) 165–172

Rb1 and Rg1 were confusing. The interaction results in Section 4.1demonstrated the existence of strong interaction between Rb1

and SSa; however, it was much weaker between Rg1 and SSaalthough Rb1 and Rg1 have similar chemical structures and similarself-assembly behavior in aqueous solution. What could be a plau-sible explanation for that phenomenon? Furthermore, it was foundthat SSa molecules had similar self-assembly behavior with Rb1

and Rg1 (Fig. 2b–d) although SSa also is a kind of saponin withan amphipathic structure. The question whether SSa moleculestend to self-aggregate in the aqueous solution or aggregate withRb1 or Rg1, needed to be addressed with more direct evidence.Therefore, the interaction parameters and energy of the blendingsystem of ginsenoside and SSa were studied to obtain a detailedunderstanding of the interaction systems.

4.3.1. Energy analysis results of ginsenoside and SSa blending systemsAs discussed in simulation section, the Flory–Huggins v and

mixing energy Emix can reflect the compatibility of different com-pounds. The compatibility is defined as the miscibility and/or de-

Table 1Flory–Huggins parameter (v), mixing energy (Emix), and potential energy of saponin/SSa b

v Emix (kcal/mol) Potential e

Electrostat

Ro/SSa �123.6480 �73.2227 �14.4588Rb1/SSa �9.4164 �5.573 �4.8203Rg1/SSa 8.5010 5.0342 �1.3056SSa/SSa 0.0000 0.0000 �2.0248

gree of interaction between different compounds [33]. The lowerv and Emix suggest better compatibility, which could result in low-er free energy. Free energy allows the judgment of the spontaneousprocess. If the free energy of a system is negative, the interaction isfavored by the components and the system would release energy.The results shown in Table 1 indicate that the order of v and Emix

values for blending systems is Ro/SSa < Rb1/SSa < Rg1/SSa. This re-sult indicates that the compatibility between different saponinswith SSa is Ro > Rb1 > Rg1. This order is also consistent with theinteraction strengths of ginsenoside and SSa. Good compatibilityof Ro and SSa also contributes toward a much lower free energyof their blending system (Fig. 3). As a result, it was easier for Roand SSa to interact spontaneously to reach a thermodynamicallystable state and to form an isotropic solution. It can be inferredfrom Fig. 3 that Rb1 could also interact with SSa spontaneously,but the interaction was relatively weaker than for Ro. The free en-ergy was positive for Rg1/SSa system, indicating the interactionwith Rg1 was not favored for SSa molecules. However, it was foundthat the free energy value was not highly positive and could be

lending system.

nergy of lowest energy configuration (kcal/mol)

ic energy Van der Waals energy Total potential energy

�6.50576 �20.9646�11.6266 �16.4469�12.4918 �13.7974�11.4733 �13.4981

Fig. 3. Free energy of ginsenoside/SSa blending system at different mole fraction ofginsenoside.

X. Dai et al. / Journal of Colloid and Interface Science 396 (2013) 165–172 169

gradually reduced to nearly zero when the mole fraction of SSa wasincreased from 0.5 to 1.0. This result indicates that the addition of alarge amount of SSa to Rg1 solution could increase the interactionprobability of SSa and Rg1. It is very plausible that this may verywell be the reason for the small interaction between SSa and Rg1,as indicated in Fig. 1b (the mole fraction of SSa in this case wasabout 0.75).

4.3.2. Interaction configurations of ginsenoside and SSa moleculesThe interaction configurations of ginsenoside and SSa with the

lowest energy are shown in Fig. 4. It can be inferred from this fig-ure that Ro interacts with SSa when its hydrophilic sugar chain,linked at C-28, interacts with the sugar chain of SSa, while itshydrophobic aglycone interacts with that of the SSa molecule(Fig. 4a). Glucuronic acid, which can ionize in aqueous medium, re-sults in much lower electrostatic energy (Table 1). The electrostaticenergy predominantly contributes to the lower total potential en-ergy of Ro and SSa molecules, which may essentially be the reasonfor the much lower free energy observed. For Rb1 and Rg1 mole-cules, the sugar chain linked at C-20 interacts with the sugar chainof SSa molecule, and another sugar chain (linked at C-3 of Rb1 mol-ecules and at C-6 of Rg1) interacts with the hydroxyl group linkedat C-24 of SSa molecules (Fig. 4b and c). Their hydrophobic struc-tures interacted with that of SSa, which was similar to Ro. How-ever, the predominant energy was Van der Waals energy, andtheir total potential energy was higher than that of Ro/SSa system,

Fig. 4. Interaction configurations of ginsenoside and SSa with lowest energy: (a) Ro/SSmolecules, and the molecules with atoms and bonds rendered as lines are ginsenoside m

resulting in higher free energy (Table 1). Rb1 molecules have moresugars, which would lead to greater number of hydrogen bondsthat can be formed between Rb1 and SSa molecules; the combina-tion was obviously stronger than for Rg1 and SSa molecules. Fur-thermore, the combination mode can also affect interactionstrength. As shown in Fig. S3 in supporting information, the twosugar chains of Rb1 can interact with the sugar chain of SSa atthe same time, while those of Rg1 could not because of limitationin the length of sugar chain. Hence, the interaction between Rg1

and SSa is much weaker than Rb1 and SSa.

4.4. Interaction of Rb1/SSa blending systems

The results and discussions above clearly indicate that Rb1 mol-ecules can interact with SSa molecules. It was of interest to studythe interaction process and its mechanism. To investigate theinteraction more clearly, and to show visual and dynamic move-ments of Rb1 and SSa molecules, studies were done at mesoscale(10–1000 nm) level to build a bridge between micro and macro-scale levels.

4.4.1. Monitoring the interaction process by DLSDLS was employed to monitor the interaction process of the

Rb1/SSa and Rg1/SSa blending systems (Rg1/SSa system was set ascontrol group) to provide directly visual results. Fig. 5 indicatesthat with elapse of time, Rb1/SSa blending system gradually be-came turbid (Fig. 5a1–a6). This observation suggests that SSa mol-ecules gradually penetrated into Rb1 micelles and formed largemicelles. Because of lower hydrophilicity, Rb1 could not form clearand stable solutions with excess amount of SSa molecules, unlikethe observation for the Ro/SSa system. Particle size distribution re-sults in Fig. 6 show that in the absence of added SSa, the diametersof Rb1 micelles were mainly distributed at about 5 nm and 249 nm.The larger size micelles were probably formed by the collision andaggregation of smaller ones, as shown by TEM images (Fig. 2b).After SSa was added, the average diameters of the aggregates in-creased over time to about 500 nm at 30 min (Fig. 6b) and in-creased to more than 800 nm after shaking the solution for morethan 1 h (Fig. 6c–f). The micelle sizes were much larger than thecolloidal particles (10–100 nm), indicating the appearance of sus-pension. It was also found that rather large aggregates appearedin the saturated solution of SSa (the diameter was larger than1000 nm, as shown in Fig. 6g), but the diameters of most aggre-gates were about 800 nm in Rb1 solution. That meant that it ismore likely that SSa molecules interacted with Rb1 instead ofself-aggregating. This result was consistent with the results ob-

a; (b) Rb1/SSa; (c) Rg1/SSa. The molecules with atoms rendered as spheres are SSaolecules.

Fig. 5. Photos of 1.0 mmol/L ginsenoside solution or 1.0 mmol/L ginsenoside with excess SSa added solution at different blending times: (a1) Rb1 solution at 4 h; (a2–6) Rb1/SSa solution at 0.5, 1, 2, 3, 4 h; (a7) SSa saturated solution at 4 h; (b1) Rg1 solution at 24 h; (b2–6) Rg1/SSa solution at 1, 5, 10, 16, 24 h; (b7) SSa saturated solution at 24 h.

Fig. 6. DLS results of Rb1 solution or Rb1 solution with excess SSa added at different blending time: (a) Rb1 solution at 4 h; (b–f) Rb1/SSa solution at 0.5, 1, 2, 3, 4 h; and (g) SSasaturated solution at 4 h.

170 X. Dai et al. / Journal of Colloid and Interface Science 396 (2013) 165–172

tained for energy analysis. All these results indicated that the Rb1

have dispersion effect on SSa. However, the Rg1/SSa solutionsshowed almost no changes within 24 h (Fig. 5b1–b6). The particlesize distributions were always 2 peaks, with one at about 100 nmand another at about 1000 nm (Fig. S4b–f in supporting informa-tion). The larger size of particles may be formed by the aggregationof small size of particles. The peak intensity for all particles wasweak, indicating only few aggregates in the solution. These resultsprove that interaction between Rg1 and SSa molecules was not sig-

nificant. It was found that the sizes were slightly different betweenthe DLS and TEM results, especially for Rb1 solutions. This wasprobably because while DLS results were the statistical averagesof the entire solution, TEM only caught a small region of thesolution.

4.4.2. Aggregation morphology study by TEMIt has been known that Ro interacts with SSa in aqueous solu-

tion to form mixed vesicles. The hydration and repulsion caused

X. Dai et al. / Journal of Colloid and Interface Science 396 (2013) 165–172 171

by glucuronic acid on the vesicle surfaces stabilizes the vesicles;this is considered as the solubilization mechanism of Ro in SSa(Fig. 7a) [23]. However, what is the interaction situation betweenRb1 and SSa molecules? TEM images in Fig. 7b show that largenumbers of spherical and worm-like micelles were formed byRb1 and SSa in aqueous solution (more TEM images are shown inFig. S5 in supporting information). The aggregation morphologieswere similar to those of quinoa saponin mixtures [6]. The worm-like micelles were the main morphology in the solution, withlengths of about 500–800 nm and widths of about 100 nm. Thesizes for spherical micelles were no larger than 100 nm diameter.

4.4.3. Study of interaction mechanism by DPD simulationThe TEM results provide a general insight into the aggregation

morphologies of Rb1 and SSa mixtures in aqueous solution. Tounderstand the interaction mechanism between Rb1 and SSa mol-ecules more clearly, DPD simulation was carried out. The coarse-

Fig. 7. TEM images of ginsenoside–SSa aggregates: (a) R

Fig. 8. Mesostructures and density distribution of Rb1/SSa aqueous solution in DPD simmesostructure of Rb1/SSa in aqueous solution; (1) section views and (2) density distribu

grain models used in this work, and their interaction parameters,are shown in Fig. S6 and Table S1 in supporting information.

The simulation results directly show the construction of the mi-celles (Fig. 8). Both spherical and worm-like micelles were ob-served to be formed by Rb1 and SSa molecules (Fig. 8c); this wasconsistent with TEM results. For spherical micelles, the hydropho-bic beads N of SSa molecules were located in, or were close to themicelle core with hydrophilic beads G and F located on the surface.Rb1 molecules were located close to the surface. Hydrophilic beadsat both ends of the molecules, which were located on the micellesurface, interact with the hydrophilic beads of SSa and separatethe hydrophobic core (formed by the hydrophobic structure ofRb1 and SSa molecules) from the water (Fig. 8d1 and d2). This kindof configuration has been found to have low energy, as discussed inSection 4.3.2. Since the Rb1 molecules were located on the micellesurfaces, their high surface activities could reduce the surface ten-sion between the SSa aggregates and water, resulting in reduction

o–SSa mixed vesicles. (b) Rb1–SSa mixed micelles.

ulation: (a) coarse-grained molecule of Rb1; (b) coarse-grained molecule of SSa; (c)tion of (d) spherical micelle and (e) worm-like micelle.

172 X. Dai et al. / Journal of Colloid and Interface Science 396 (2013) 165–172

of pressure (which can reflect the tendency of free energy tochange) for Rb1/SSa/water blending system. As a result, the spher-ical micelle would enlarge to form worm-like micelles. The forma-tion process indicated in the simulation results was found throughthe fusion and modification of small spherical micelles. (The pres-sure changes and detailed simulation results are shown in Fig. S7 insupporting information.) The worm-like micelles were not simplecylindrical, as shown in Fig. 8c and e1. The section view indicatesthat they were a little compressed. The radius in minor axis direc-tion was nearly equal to the length of the extended hydrophobicstructure of SSa molecule. The density distribution of the axialdirection of the section view indicates that SSa molecules occupiedthe micelle core, with Rb1 molecules adsorbed close to the surface(Fig. 8e2). The situation was presumably similar with spherical mi-celles. The hydrophilic beads for both SSa and Rb1 molecules lo-cated on the surface formed a hydrophilic layer and stabilizedthe micelles through hydration.

5. Conclusion

In this paper, ginsenosides Ro, Rb1, and Rg1 were found to havedifferent interaction effects on SSa in aqueous solution: Ro couldmarkedly increase the solubility of SSa; Rb1 could disperse SSa inaqueous medium; Rg1 had no significant interaction with SSa.The understanding of differences in the interactions of ginsenosidewith SSa or other similar compounds could be of immense benefitin classifying different ginsenosides directed at specificapplications.

To understand the interaction mechanism of ginsenoside withSSa more clearly, a combination of multiscale methods was used.It was found that the chemical structure, especially the type andnumber of hydrophilic sugars, has significant influence on theinteraction effect. Glucuronic acid, which could significantly re-duce the potential energy through its strong electronic interactionwith the sugars of SSa, contributed significantly to the high com-patibility between Ro and SSa molecules. This resulted in a muchlower free energy of Ro and SSa blending system, and the sponta-neous interaction between the two components took place moreeasily. In comparison with Rg1, greater number of sugars in Rb1 re-sulted in more binding sites with SSa. The sugars could also in-crease the stability of the mixed micelles formed by Rb1 and SSamolecules, resulting in a stronger interaction between Rb1 andSSa than between Rg1 and SSa. However, the answer to the ques-tion, if more sugars always indicate stronger interactions, still re-quires more evidence.

Studies on aggregation morphologies of ginsenoside showedthat Ro could form vesicles in aqueous solution, while Rb1 andRg1 formed spherical micelles. The opportunities for SSa moleculesto interact with aggregates were also different because of differ-ences in the available surface area; these differences could havecontributed, to some extent, to the different interaction effects.Spherical and worm-like micelles were found to be formed byRb1 and SSa molecules. This behavior is different from Ro andSSa which formed vesicles [23], but similar to those formed by qui-noa saponin mixtures [6]. The formation appears to go through thefusion and modification of small spherical micelles. These resultsare significant because such properties, and applications of sapo-nins often associated with their aggregation, can help developnew applications of ginsenosides.

Acknowledgments

This work was financially supported by the National NaturalScience Foundation of China (81073058) and Innovation TeamFoundation of Beijing University of Chinese Medicine, Beijing(2011-CXTD-11, Research Center of TCM-informationEngineering).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.jcis.2013.01.017.

References

[1] Ö. Güçlü-Üstündag, G. Mazza, Critical Reviews in Food Science and Nutrition47 (2007) 231.

[2] A. Osbourn, R.J.M. Goss, R.A. Field, Natural Product Reports 28 (2011) 1261.[3] U.A. Emirates, Safety Evaluation of Certain Food Additives and Contaminants

(2004) 639.[4] K. Dalsgaard, Archiv für die gesamte Virusforschung 44 (1974) 243.[5] J.C. Cox, A.R. Coulter, B. Morein, K. Lovgren-Bengtsson, B. Sundquist, Google

Patents, 2002.[6] S. Verza, P. de Resende, S. Kaiser, L. Quirici, H. Teixeira, G. Gosmann, F. Ferreira,

G. Ortega, Die Pharmazie – An International Journal of Pharmaceutical Sciences67 (2012) 288.

[7] C.A. Kensil, D.J. Marciani, Google Patents, 1991.[8] B.P. Chapagain, Z. Wiesman, Journal of Agricultural and Food Chemistry 54

(2006) 6277.[9] O. Tanaka, N. Yata, Google Patents, 1985.

[10] S. Tsuzaki, S. Wanezaki, H. Araki, Google Patents, 2008.[11] H. Kimata, T. Nakashima, S. Kokubun, K. Nakayama, Y. Mitoma, T. Kitahara, N.

Yata, O. Tanaka, Chemical and Pharmaceutical Bulletin 31 (1983) 1998.[12] H. Kimata, N. Sumida, N. Matsufuji, T. Morita, K. Ito, N. Yata, O. Tanaka,

Chemical and Pharmaceutical Bulletin 33 (1985) 2849.[13] M. Toshinobu, N. Rui-lin, K. Hiroko, I. Keiko, M. Noriko, K. Ryoji, H.J. Zhou, W.

Cheng-yih, Y. Noboru, O. Tanaka, Chemical and Pharmaceutical Bulletin 34(1986) 401.

[14] Y. Sasaki, K. Mizutani, R. Kasai, O. Tanaka, Chemical and PharmaceuticalBulletin 36 (1988) 3491.

[15] K. Watanabe, H. Fujino, T. Morita, R. Kasai, O. Tanaka, Planta Medica 54 (1988)405.

[16] X.H. Zhou, R. Kasai, M. Yoshikawa, I. Kitagawa, O. Tanaka, Chemical andPharmaceutical Bulletin 39 (1991) 1250.

[17] U. Walthelm, K. Dittrich, G. Gelbrich, T. Schoepke, Planta Medica 67 (2001) 49.[18] M. Kubo, T. Tani, T. Katsuki, K. Ishizaki, S. Arichi, Journal of Natural Products 43

(1980) 278.[19] Y. Nagai, O. Tanaka, S. Shibata, Tetrahedron 27 (1971) 881.[20] Q. Du, G. Jerz, R. Waibel, P. Winterhalter, Journal of Chromatography A 1008

(2003) 173.[21] O. Tanaka, S. Yahara, Phytochemistry 17 (1978) 1353.[22] R.M. Corbit, J.F.S. Ferreira, S.D. Ebbs, L.L. Murphy, Journal of Agricultural and

Food Chemistry 53 (2005) 9867.[23] X. Dai, X. Shi, Y. Wang, Y. Qiao, Journal of Colloid and Interface Science (2012).[24] S. Yamamoto, Y. Maruyama, S. Hyodo, The Journal of Chemical Physics 116

(2002) 5842.[25] X.D. Guo, L.J. Zhang, Z.M. Wu, Y. Qian, Macromolecules (2010).[26] C. Zhong, D. Liu, Macromolecular Theory and Simulations 16 (2007) 141.[27] X. Guo, L. Zhang, Y. Qian, J. Zhou, Chemical Engineering Journal 131 (2007) 195.[28] J. Xia, C. Zhong, Macromolecular Rapid Communications 27 (2006) 1654.[29] X.D. Guo, J.P.K. Tan, L.J. Zhang, M. Khan, S.Q. Liu, Y.Y. Yang, Y. Qian, Chemical

Physics Letters 473 (2009) 336.[30] L.S. Zheng, Y.Q. Yang, X.D. Guo, Y. Sun, Y. Qian, L.J. Zhang, Journal of Colloid and

Interface Science 363 (2011) 114.[31] C. Soto-Figueroa, L. Vicente, Soft Matter 7 (2011) 8224.[32] P. Posocco, M. Fermeglia, S. Pricl, Journal of Materials Chemistry 20 (2010)

7742.[33] X.D. Guo, Y. Qian, C.Y. Zhang, S.Y. Nie, L.J. Zhang, Soft Matter (2012) 9989.[34] F. Case, P. Alexandridis, Mesoscale phenomena in fluid systems. ACS

Publications, 2003.[35] R.D. Groot, P.B. Warren, The Journal of Chemical Physics 107 (1997) 4423.[36] K. Holmberg, B. Jönsson, B. Kronberg, B. Lindman, Wiley Online, Library, 2004.[37] D. Myers, Surfactant Science and Technology, Wiley-VCH, 2005.