Synergistic Effect of Compounds from a Chinese Herb: Compatibility and Dose Optimization of Compounds from N-Butanol Extract of Ipomoea stolonifera

The herbal medicine Ipomoea stolonifera (I. stolonifera) has previously been shown to have considerable anti-inflammatory potential in vivo and in vitro. To establish a method for exploring the synergistic effects of multiple compounds, we study the compatibility and dose optimization of compounds isolated from n-butanol extract of I. stolonifera (BE-IS). Raw264.7 cell was treated with lipopolysaccharide (LPS) in the presence of compounds from BE-IS, namely scopoletin, umbelliferone, esculetin, hesperetin and curcumin, using the orthogonal design, uniform design and median-effect method. To verify the best efficacy of principal constituents in vivo, the uniform design was used in the croton oil-induced mouse ear edema model. The results from LPS-induced the production of prostaglandin E2 (PGE2) show that, esculetin, curcumin and hesperetin were the principal constituents that had synergistic effects when used at the optimal ratio. Additionally, the principal constituents were found to work synergistically in the croton oil-induced mouse ear edema model at low doses. It turned out that the three experimental optimization and analysis methods (orthogonal design, uniform design and median-effect method) can be effectively used to solve both compatibility and dose optimization for combined use of multiple compounds.

Screening principal constituents from BE-IS by orthogonal design. According to the effective concentration range of compounds from BE-IS, 4 different concentrations were chosen for them individually. The L 16 (4 5 ) orthogonal design table (Table 2) was used to design an experiment comprising 16 combinations. Direct analysis and variance analysis were used to analyze the result, confirming the influence degree of each compound on the combinative effect.
In Table 2, K1, K2, K3, and K4 indicate the sum of the inhibition rates of each factor at various levels; avK1, avK2, avK3, and avK4 indicate the average inhibition rate of each factor at various levels; and R is the range of the average inhibition rate of each factor at various levels (range = maximum of average inhibition rate -minimum of average inhibition rate). It reflects the influence on the experimental result by variations in the level of each factor. According to R value and analysis of variance (Table 2), variation in the concentration of Esc had the greatest effect on the result (P < 0.01), followed by variations in the concentration of Cur (P < 0.01) and Hes (P < 0.05), respectively. Therefore, these were determined to be the principal constituents from BE-IS when used in combination.
The optimal concentrations and best efficacy of principal constituent analysis by uniform design. In this study, Esc, Cur and Hes were found to be the principal constituents by orthogonal design.
Seven different concentrations were chosen for Esc, Cur and Hes, individually. The U 7 (7 3 ) uniform design table (Table 3) was used to design an experiment of 7 combinations (Table 3). Esc, Cur and Hes were independent variables, and the PGE 2 release inhibition rate was the dependent variable Y. Excel Solver tool was used to find the minimum values of the residual sum of squares between the experimental values and the predicted values, and Scientific RepoRts | 6:27014 | DOI: 10.1038/srep27014 a multivariate quadratic equation was established: Y = − 0.3495 × E 2 + 8.0821 × E + 0.1480 × C 2 + 2.9069 × C − 0.2212 × H 2 + 5.5140 × H − 16.7161. The data indicated that Esc and Hes had a quadratic response surface effect. Additionally, it was found that the efficacy of Cur increased in direct correlation to an increase in concentration. Because figures are not adequate to illustrate multi-dimensional surfaces, a response surface figure was drawn with a fixed Cur, where Esc and Hes were represented on the X and Y axis, respectively, and PGE 2 release inhibition rate was represented by the Z axis (Fig. 2).
The above equation and response surface analysis predicted that the optimal concentrations were: Esc 11.6 μ M, Cur 8.5 μ M, and Hes 12.5 μ M with an expected efficacy of 99.77%. Because this combination was not one of the seven combinations tested, this combination was further tested in parallel with combination 7, which was the combination with the best efficacy in the seven tested combinations. These two combinations showed similar results upon testing ( Table 3), suggesting that our prediction was correct.

Interaction results among principal constituents determined by median-effect method analysis.
Each drug had 5 pairs of D and Fa ( Table 4). Correlations of log-log coordinates were examined and the correlation coefficient R was calculated. Double logarithmic coordinates linear regression analysis was conducted by using equation (1), and m and D 50 were calculated (Table 5). R was found to be 0.99 for each drug, suggesting that our experimental design was appropriate. When D 50 and m were obtained, the function relation between dose D and inhibition rate Fa was established by using equation (1).
As shown in Table 5, the m values of Esc, Cur, and Hes, and combinations of Esc & Cur, Esc & Hes, and Cur & Hes were similar. Their double logarithm coordinate lines were almost parallel, indicating that the effects of these compounds were not independent, and therefore k = 0. Assuming Fa = a%, the combination index (CI)a was obtained by equations (1), (2), (3), and (4). As shown in Table 5, when combination of Esc & Cur and Cur & Hes were given at high or low dosages, CI < 1; when combination of Esc & Hes had Fa ≥ 30%, CI < 1.   Effect of principal constituents from BE-IS on the production of NO and cytokines. The data indicated that Cur effectively inhibited LPS-induced the production of NO (IC 50 = 15.08 μ M) and IL-1β (IC 50 = 33.69 μ M), whereas Esc inhibited LPS-induced the production of IL-6 (IC 50 = 28.29 μ M), but had no effect on TNF-α or IL-1β . Additionally, Hes inhibited the production of NO, TNF-α , IL-1β and IL-6, but not significantly (Fig. 3).
The optimal dosages and best efficacy of principal constituent analysis by uniform design in vivo. Seven different dosages were chosen for each of the principle components and the U 7 (7 3 ) uniform design table (Table 6) was used to design an experiment of 7 combinations (Table 6). In Table 6, Esc, Cur, and Hes  were independent variables, and inhibition of croton oil-induced mouse ear edema was the dependent variable Y. Excel Solver tool was used to find the minimum values of residual sum of squares between the experimental values and the predicted values, and a multivariate quadratic equation was established: The above equation and Excel Solver tool predicted that the optimal dosages were: Esc 337 mg·kg −1 , Cur 60 mg·kg −1 , and Hes 191 mg·kg −1 . The expected efficacy was: 99.63%. Because this combination was not one of the 7 combinations tested, this combination was further tested in parallel with combination 6, which was the combination with the best efficacy in the seven tested combinations. These two combinations showed similar results   upon testing ( Table 6), suggesting that our prediction was correct. Results indicate that combination therapy may prove invaluable given the synergistic potential as well as the possibility of utilizing smaller dosages.
Verification of the results. To verify the results above, we compared the combination to each compound of Esc, Cur and Her at the indicated optimal concentrations or dosages (Esc 11.6 μ M, Cur 8.5 μ M, and Hes 12.5 μ M in vitro and Esc 337 mg·kg −1 , Cur 60 mg·kg −1 , and Hes 191 mg·kg −1 in vivo.) on LPS-induced the production of the inflammatory mediators and cytokines in vitro, and croton-oil induced mouse ear edema in vivo. The results showed that the combination of Esc, Cur and Her at the indicated optimal concentrations did not have any significant cytotoxic effects on Raw264.7 cells and inhibited the production of NO, PGE 2 , TNF-α , IL-1β and IL-6 in vitro and decreased ear edema in vivo significantly (Fig. 4).

Discussion
I. stolonifera has been consumed for many years and touted for its anti-inflammatory properties. Relatively recent research has identified a number of compounds, isolated in n-butanol, that could be responsible for the anti-inflammatory properties of the herb. Despite their discovery and the implications for the treatment of inflammation, how these compounds interact, as well as their mechanisms of action are still unclear. Therefore, research into the synergistic potential between compounds as well as dosage optimization is a vital area of research.
Single herbs with multiple constituents can have multiple targets, suggesting that efficacy can be achieved by the synergistic and dynamic interactions among the multiple constituents. Overall quantitative analysis of drug combinations often involves multiple factors and multiple levels. Such studies mostly use orthogonal design [17][18][19] and the orthogonal t value method 20 . These experiments can identify primary and secondary factors, but cannot be used to find the optimal concentrations of drugs in combination when there are many dosage levels. An improved uniform design method 21-26 is available for multiple drugs combination analysis to determine optimal concentrations of drugs, especially the constituents of Chinese complex prescriptions 27 . This study represents the first attempt at applying the orthogonal design in order to identify primary and secondary interaction of compounds isolated from I. stolonifera. Initially, we sought to prove that synergistic effects existed between principal constituents at any ratio, by using the median-effect method, and then use the uniform design method to find the optimal concentrations and best efficacy of principal constituents for use in combination. Results of preliminary investigations indicated that synergistic interactions between principal constituents at any ratio may not be obvious in most situations. Thus we first identified the optimal concentrations of principal constituents by uniform design, and then using the median-effect method, we found that synergistic effects clearly exist between the principal constituents when they are used at the optimal concentrations.
Compounds isolated from I. stolonifera were found to inhibit LPS-induced PGE 2 release, a widely used inflammation marker 28,29 . Collected data indicated that the inhibition of cyclooxygenase (COX)-influenced arachidonic acid metabolism may result in a subsequent reduction in PGE 2 synthesis and inhibition of the inflammatory response. Therefore, the model of LPS-induced PGE 2 release in Raw264.7 was adopted to evaluate the combined effect of compounds from I. stolonifera. Direct analysis and variance analysis were used to analyze the orthogonal experiment results 30 . The former is direct and easy to use, and has a low computing work load, but it cannot estimate experimental error. Namely, it can not distinguish whether the cause of experimental result variation is either due to variations in the tested factors or random. Analysis of variance can solve this problem. For analysis of variance, random error should be estimated and obtained from the blank column in the orthogonal chart. To reduce experiment times, the minimized sum of squares of deviations was chosen as an approximate evaluation. Results of both analyses indicated that the compounds that significantly affected release of PGE 2 are Esc (P < 0.01), Cur (P < 0.01), and Hes (P < 0.05). Thus, they were identified to be the principal constituents from BE-IS. In uniform design experimental data processing, all factors should be used for regression analysis. The purpose of this is to obtain the optimal concentrations or dosages through partial derivatives. The expected best efficacy was found by response surface analysis, and at the optimal concentrations, the median-effect method was used to analyze the interaction among principal constituents. The median-effect method, namely Chou-Talalay combination index method 31 , is deduced by Chou and Talalay from the median-effect equation based on the law of mass action. As a mechanism-based, combined action analytical method, it has been one of the most commonly used methods. The results show that the principal constituents have synergistic effects when used at the optimal concentrations, suggesting that the anti-inflammatory effect of BE-IS is caused by synergistic actions of the individual compounds.
Data indicate the distinct abilities of the identified compounds to inhibit inflammatory mediators and cytokines. For example, Cur inhibits LPS-induced the production of NO and IL-1β most significantly, whereas Esc inhibits IL-6, but has no effect on LPS-induced the production of TNF-α and IL-1β . We speculate they probably act during different steps of inflammation, thus the combined use of these individual compounds can exert synergistic anti-inflammatory effects. We find that combined use of principal constituents from BE-IS can improve treatment efficacy while reducing the dosages. The verification results futher demonstrate that this combination is indeed synergistic, which could inhibit the production of inflammatory mediators and cytokines more significantly than sigle use of the compounds at the indicated optimal concentrations.
The three compounds (Esc, Cur and Her) exist in the highest quantity in the extract compared to the other bioactive compounds in the herb. Like the n-butanol extract, chloroform, acetic ether and aqueous extracts of I. stolonifera probably have the similar results that Esc, Cur and Her still have synergistic effects in the extracts. Though the three compounds show synergistic effects on the inhibiton of PGE 2 release, they would not so sure have synergistic effects if NO, TNF-a, IL-1β or IL-6 levels were used as the parameter for the basis of the orthogonal design, uniform design and median effect analysis. In our study, the combination of Esc, Cur and Her at the indicated optimal concentration inhibits NO, TNF-a, IL-1β or IL-6 levels effectively. Based on the results, the effects of combining the three compounds are as good as using whole n-butanol extract. Obviously, the combination has more advantages, such as the convenient and abundant resource, and the certainty composition. We suggest that using the combination of the three compounds instead of whole herbal extract for its anti-inflammatory function. That is just why we study the compatibility and dose optimization of mutiple compounds from the herb.

Conclusions
This study provides a new mode for compatibility analysis and concentration optimization of compound preparations of Chinese medicine. The mode includes screening principal constituents by orthogonal experiment, analyzing the optimal concentrations and best efficacy of principal constituents by uniform design, determining the synergistic effects between principal constituents by the median-effect method and investigating the mechanisms of these synergistic effects.

Methods
Animals. Male 4-week-old Kunming mice kept under SPF (specific-pathogen-free) conditions were obtained from the Laboratory Animal Center, Shantou University Medical College (Shantou, China). Animals were Scientific RepoRts | 6:27014 | DOI: 10.1038/srep27014 maintained in a room at a controlled temperature of 23 ± 2 °C and a fixed 12h light/dark cycle with free access to food and water. The total number of animals used in this study was 182 for the croton oil-induced mouse ear edema model. All studies involving animals are reported in accordance with the ARRIVE guidelines for reporting experiments involving animals 32,33 . All protocols involving animals were approved by the Institutional Animal Care and Use Committee of Shantou University Medical College.
Cell culture and reagents. The  MTT assay. The cytotoxicity of compounds from BE-IS was determined by MTT assay 34 . Raw264.7 cells (~1 × 10 5 cells·mL −1 ) were plated onto 96-well culture plates and incubated overnight. The cells were untreated or treated with 1 μ g·mL −1 LPS, 0.1% DMSO, BE-IS compounds (dissolved in 0.1% DMSO) or 1 h treatment of BE-IS compounds, followed by 1 μ g·mL −1 LPS, and then incubated at 37 °C under a humidified atmosphere with 5% CO 2 for 24 h. Cells were then incubated with MTT for 4 h, followed by the addition of 150 μ L DMSO per well. After solubilizing completely, cytotoxicity was determined by measuring the OD at 550 nm using a microplate reader.
Measurement of PGE 2 , NO and cytokines. Measurement of cytokines, NO and PGE 2 was performed as previously described 7 . Briefly, Raw264.7 cells were plated ~2 × 10 5 cells·mL −1 per well in 24-well culture plates. Twenty-four hours later, cells were treated with BE-IS compounds, individually or combined, for one hour, before stimulation with 1 μ g·mL −1 LPS for 24 h. The supernatant was collected and the amount of secreted PGE 2 and cytokines was measured by specific ELISA kits (PGE 2 , Enzo Life Sciences Inc., NY, USA; TNF-α , IL-1β and IL-6, Boster Bioengineering Ltd, Wuhan, China) according to the manufacturer's instructions. The level of NO in the culture supernatant was measured using an NO assay kit (Jiancheng Bioengineering Institute, Nanjing, China) according to the manufacturer's instructions.
To study their dose-efficacy relationship of single compounds on inhibiting PGE 2 release, the concentration-effect relationship of each compound from BE-IS was fitted to the Hill equation y = E max × x n / (k n + x n ), using Origin7.5 software.
Orthogonal experimental design. Based on the effective concentration range of compounds from BE-IS, the L 16 (4 5 ) orthogonal design table 17 was used to design an experiment with 4 levels per compound and a total of 16 combinations. Inhibition of LPS-induced PGE 2 release was used as an indicator to evaluate the efficacy. Through direct analysis and variance analysis, the effect of each compound on the efficacy of the tested combinations was determined to screen the principal constituents in BE-IS.
Uniform experimental design. Based on the results from orthogonal experimental design, the U 7 (7 3 ) uniform design table 23,27 was used to design an experiment with 7 different combinations. Inhibition of LPS-induced PGE 2 release was used as an indicator to evaluate the efficacy. A multivariate regression equation was established using Excel Solver tool, and optimal concentrations, as well as efficacy of combined use of principal constituents from BE-IS were determined by the response surface method.
Median-effect method. Based on the results from uniform experimental design, the proportion for the combined use of the principal constituents was set. The median-effect method was used to analyze the interaction among principal constituents by using the following equation: Fa/Fu = (D/D 50 ) m , where Fa denotes efficacy, Fu = 1-Fa, D indicates drug dosage, m is the parameter, and D 50 refers to the 50% efficacy dosage. In order to account for the variable D 50 of the drugs and m, the logarithm was taken as: log (Fa/Fu) = mlogD-mlogD 50 (1). For each drug, dosage D and the corresponding inhibition rates Fa were determined from the experiment. The D 50 and Fa were determined from equation (1) by doubling the logarithm coordinates and performing a linear regression analysis Assuming the test drugs were Drug 1, Drug 2, and their combination (Drug 1, 2), when inhibition rate Fa = a%, the required dosage (Da)1, (Da)2, and (Da)1, 2 could be calculated from equation (1). Assuming the ratio of drug 1 to drug 2 in Drug 1, 2 was a:b, the dosage of drug 1 and drug 2 in (Da)1, 2 can be calculated by: (da)1 = (Da)1, 2* a/(a + b) (2); (da)2 = (Da)1, 2* b/(a + b) (3). The combination index (CI), which indicates the efficacy of drugs used in combination, can be obtained based on the median-effect method: (CI)a = (da)1/ (Da)1 + (da)2/(Da)2 + k* (da)1(da)2/(Da)1(Da)2 (4), where k refers to the parameter for combined action. When the effects of two drugs are not independent, k = 0. When the effects of two drugs are independent, k = 1. Therefore, a CI < 1 indicates synergy, CI = 1 indicates additivity, and CI > 1 indicates an antagonistic effect between the two drugs.
Croton oil-induced ear edema model. Croton oil-induced ear edema was performed as described previously 35,36 . Mice were divided randomly into the following eight groups (12 mice per group) for oral gavage treatment once a day for 5 days: control group (0.5% CMC-Na, 0.2 ml) and seven treated groups (the combination of different doses of Esc, Cur and Hes). One hour after the last drug administration, ear edema was induced by topical application of 1% croton oil in mixture solvent (v/v, croton oil: ethanol: pyridine: ethyl ether = 1:10:20:69, 100 ul) on the outer and inner surfaces of the right ear of each mouse. The left ear remained untreated and served as a control. The mice were killed by cervical dislocation 2 h after the application of the irritant. An ear disk, Scientific RepoRts | 6:27014 | DOI: 10.1038/srep27014 8.0 mm in diameter, was punched out and weighed. The weight difference between the left and the right ear disk of the same animal was evaluated as the extent of edema. The inhibition percentage was calculated by the following equation: where E control and E treated is the extent of edema from the control group and treated groups.
Uniform design in vivo experiment. The U 7 (7 3 ) uniform design table was used to design an experiment with 7 different combinations. Inhibition of croton oil-induced edema in the mouse ear was used as an indicator to evaluate the efficacy. A multivariate regression equation was established, and optimal concentrations as well as efficacy of combined use of principal constituents from BE-IS, were determined with Excel Solver tool.
Verification experiments. According to the results above, the optimal concentrations or dosages of principal constituents were: Esc 11.6 μ M, Cur 8.5 μ M, and Hes 12.5 μ M in vitro and Esc 337 mg·kg −1 , Cur 60 mg·kg −1 , and Hes 191 mg·kg −1 in vivo. To prove the results, the experiments below were designed. MTT assay was performed to determine whether the combination of Esc 11.6 μ M, Cur 8.5 μ M, and Hes 12.5 μ M affects viability of Raw264.7 cells. In LPS-induced Raw264.7 cells, NO assay kits and specific ELISA kits were used to determine the levels of NO, PGE 2 , TNF-α , IL-1β and IL-6 after treatment with the combination and each compound of Esc, Cur and Her at the indicated optimal concentrations. Croton oil-induced mouse ear edema was performed to determine that the inhibition of ear edema after treatment with the combination and each compound of Esc, Cur and Her at the indicated optimal dosages. Statistical analysis. The experimental data were analyzed with the SPSS 17.0 software and presented as the mean ± standard error (x ± S.E.). Comparison between groups was performed using one-way ANOVA, and results were considered statistically significant when P < 0.05.