An alternative green separation process for the pure isolation of commercially important bioactive molecules from plants

2.1 Chemicals and plant material

Arnebia nobilis (root bark), Citrus paradisi (fruit rind), and Asparagus officinalis (aerial parts) were obtained from local market in Thanjavur, India. Cotoneaster buxifolia (aerial parts), Ventilago maderasapatana (roots), and Oroxylum indicum (leaves) were obtained from ornamental farms of Nilgiris, Tamil Nadu. The plant was botanically authenticated by Dr. Jayendran, Department of Botany, Government Arts College, Ootacamund, India. A voucher specimen of all plant samples was deposited in Government Arts College, Ootacamund, India. The plant materials were shade dried and ground to semi-coarse powder and directly used for extraction process. Sucrose, sodium hydroxide, and hydrochloric acid were purchased from Merck Ltd., India. Pure molecules (standards) were obtained from Tokyo chemical industry, Japan.

2.2 Analytical methods

Purity of the product obtained from extraction experiments was analyzed by high performance liquid chromatography (HPLC-DAD). Commercial grade of compounds was used as standards for the experiments. An Agilent 1200 Series HPLC system (Waldbronn, Germany) with a G1315D diode-array detector was used. Samples (20 μl) were injected into Phenomenex Kinetex XB-C18 column (75 mm×3 mm, 2.6 μm). The mobile phase consisted of 10 mm ammonium formate (NH 4COOH) in water:acetonitrile (98:2) (A) and 10 mm ammonium formate (NH 4COOH) in water:acetonitrile (2:98) (B). The flow rate was maintained at 1.0 ml/min. The HPLC analysis was done at each step of developed process to determine the amount of flavonoids. The extracted precipitate from A. nobilis was analyzed by LC-MS on an Agilent LC-MS system using Acentis express C18 column (50×2.1 mm; 2.7 μm) using electrospray ionization as the ion source (spray voltage of 3.5 kV, flow rate of 0.1 ml min⁻¹). ¹H NMR, ¹³C NMR, and DEPT-135 NMR spectra were determined on a Bruker-300 NMR spectrometer, and chemical shifts were expressed as parts per million against TMS as internal reference.

2.3 Solubility studies: alkaline sucrose

Initially, solubility of Pure BDMS and UA was studied in various concentrations of sucrose solutions. Later, the study was extended to other molecules (Figure 1). The experiments were done in a glass vessel (cylindrical) (50 ml). Pure compounds were suspended in these vessels containing alkaline sucrose solutions at various concentrations (20%–70%). Alkaline sucrose solution was prepared by the addition of sodium hydroxide solution (20%) to attain pH of 12–14. Excess amount of compounds was equilibrated in these solutions and kept under vigorous stirring for 48 h at 35°C using a magnetic bar. After 48 h, the solution was allowed to settle and the clear solution was aspirated to analyze the amount of compounds solubilized. Also, the stability of compounds at different temperatures was studied and subsequently used in defining the optimum temperature range for extraction of compounds.

Figure 1:

β,β-Dimethylacryl shikonin (1), ursolic acid (2), Physcion (3), Rutin (4), Naringin (5), and Chrysin (6).

2.4 Effect of pH on precipitation

The clear solution aspirated from previous experiments was neutralized using dilute hydrochloric acid (2N) added drop wise. The precipitated compounds were analyzed for their weights at different pH (12–5) levels. The same experiments were also repeated during extraction of these compounds from their respective plant matrices. The purity and yield were analyzed using HPLC.

2.5 Alkaline sucrose based extraction

For extraction of BDMS, A. nobilis (root bark) was suspended in a glass vessel (cylindrical) (500 ml) with baffles and fitted with vacuum vent. Initial vacuum of –20 (in Hg) was applied using diaphragm vacuum pump (Riviera^®). The concentration of sucrose varied based on optimization experiments. pH was maintained as 12 throughout the tests. The mixture was vigorously agitated (1000 rpm) for 8 h using cross-spin magnetic stirring bar. Sample aliquots at definite time intervals were taken and analyzed for the BDMS content. At the end of extraction, the clear solution containing the metabolites was filtered under vacuum. A moderate blue color filtrate was obtained. The insoluble sticky residue was washed with 10 ml of alkaline sucrose solution, filtered, and mixed with the filtrate. The filtrate was neutralized with HCl (2 N), which afforded to produce precipitate. The precipitate was filtered, washed, and analyzed for amount and purity of BDMS. Similar extraction was done for other compounds.

2.6 Influence of extraction variables

Before implementing a standardized experimental design protocol and progression of the study by RSM based modeling, a maiden set of tests was performed by following the classical “one variable at a time” optimization approach to roughly select the applicable factors and their range in extraction process. Parameters such as concentration of sucrose (X₁), temperature (X₂), solid loading (X₃), extraction time (X₄), agitation (X₅), vacuum (X₆), pH of extraction (X₇), and pH of precipitation (X₈) were studied for their influence on final yield and purity of both BDMS and UA. The experiments were conducted separately in fully baffled glass container (50 ml) with magnetic stirrer, and the parameters were identified for the possible influence. Most of the influential experimental parameters that increased yield of compounds obtained were analyzed by HPLC and were further considered for systematic experimental design to find the optimum parameter, set through RSM procedure.

2.7 Response surface methodology

The influential parameters were identified from preliminary experiments performed by analyzing their effect on yield of BDMS and UA, respectively (target responses). Consequently, concentration of alkaline sucrose (%), temperature (°C), and solid loading (%) were the only parameters found to be effecting the responses. Therefore, three factors and three levels (–1, 0, +1) based on their scanned range were advanced for Box-Behnken method. The design generated 15 sets of experiments/runs, which were done for three replicates, and the average is represented in Table 1. To minimize the effects of inexplicable variability in the observed response due to inessential factors, the order of experiments was randomized. Experimental data thus obtained were applied into second-order polynomial model, and the regression coefficients were determined as in Equation (1).

where Y is the predicted response factor; β₀ is the intercept; and β_i,β_ii, and β_ij are regression coefficients for linear effects, regression coefficients for squared effects, and regression coefficients for interaction effects, respectively; and X_i and X_j are the parameters.

2.8 Statistical analysis

Results obtained were expressed as the mean±standard deviation of the replications. Results obtained in experimental run generated by BBD were expressed as mean of replicates. RSM based model fitting and statistical analysis was executed using Design Expert (v. 9.0.3.1; State-Ease, Inc., USA). An analysis of variance (ANOVA) was done to find the significant levels defined at p<0.05, p<0.01, and p<0.001.

Extractions were always done in replicates at all points in the design study. The corresponding extracts were checked for the dependent variables (responses): amounts of BDMS (Y₁) and UA (Y₂). To determine the relationship between the experimental levels and the response of each factor, response surfaces graphs were plotted using reduced fitted polynomial models.

3.1 Preliminary observations and analytical methods

The initial observations with A. nobilis depicted the potential of the developed process due to the formation of precipitate after neutralization. The precipitates obtained for BDMS looked pure with respect to color and texture. The color of extract and texture of precipitate for BDMS are shown in Figure 1A. The change in color of solubilized molecules is due to the introduction of a substituent at different pH, especially a hydroxyl group (either free or methylated), which may induce a bathochromic shift of the absorption band. Thus, shikonins are red at pH 7 and blue at pH 12. The extracted yields obtained were significant for any extraction process. However, as purity is of prime concern when isolating plant active molecules, HPLC-DAD analysis was done depicting a purity of 93.7% for BDMS (Figure 2B). Before, advancing to further structural characterization, a mass spectral analysis for the precipitate obtained from the A. nobilis was done, which suggested it to be BDMS (1) (Figure 2C). Furthermore, the molecule was analyzed by nuclear magnetic resonance spectrometry (NMR), which confirmed the structure to be BDMS (Figure 2D and E). For BDMS, optical rotation studies solved the uncertainty of being either an alkannin derivative or its stereoisomer shikonin. The specific rotation [α]D20 of BDMS was found to be +304 (c 0.5, CHCl₃), agreeing with the literature reports [20]. These remarkable observations led to the application trials of developed process in extraction of other molecules. The selection of plants was made so as to rigorously validate the reliability of the process, and therefore, different plant parts from different plants were collected and tested. In all the experiments, the lead molecules were precipitated by the process. On the basis of the spectral data (Supplemental Figures S1–S5), the structure of the compounds was identified and confirmed as UA (2), Physcion (3), Rutin (4), Naringin (5), and Chrysin (6) (Figure 1). All of the spectral data of isolated compounds in current study agreed with the literature [21], [22], [23], [24], [25].

Figure 2:

(A) Alkaline sucrose extract of A. nobilis (left), precipitate obtained after neutralization at pH 7 [β,β-dimethylacryl shikonin (BDMS)=93.75% pure) (right); (B) HPLC-DAD profiles of pure BDMS isolated using alkaline sucrose extraction process from A. nobilis; (C) mass spectra of BDMS, (D) proton nuclear magnetic resonance spectra of BDMS, (E) carbon nuclear magnetic resonance spectra of BDMS.

3.2 Solubility studies: alkaline sucrose

To understand the behavior of alkaline sucrose as extracting solvent, its efficiency of solubilizing the target compounds should be understood. For this, the solubility of pure compounds in different concentrations of sucrose was studied as given in Figure 3.

Figure 3:

(A) Effect of concentration of sucrose on solubility of β,β-dimethylacryl shikonin (BDMS) at different pH. (B) Ursolic acid (UA).

The pattern shows increase in solubility of both BDMS and UA with increasing concentrations of sucrose until 50%. Then the solubility decreases. The plot also shows effect of pH on solubilization, which depicts that at pH 12, the solubilization is at a maximum and then becomes stationary at higher pH. Based on these observations, 50% sucrose at pH 12 was taken for further optimization and batch extraction.

Sucrose is a special disaccharide that, in concentrated solutions, folds around its glycosidic linkage and makes intramolecular hydrogen bonds [26]. In this form, it mostly exposes its OH groups readily for bonding. When sucrose concentration in solution is high enough, there is a possibility for the existence of two different conformations: (i) intramolecular hydrogen bond (O2p-H1f) and (ii) H2p-O1f [27]. Here, p stands for glucose (pyranose) and f means furanose. Sometimes, furanose moiety rotates around the glycosidic linkages, thus forming O2p-H6f or O6f-H2p [28]. At high concentrations, the bending of sucrose molecule around the glycosidic bond exposes all of its multiple -OH groups for complex formation. De-protonation of these -OH groups happens at high pH leading to ionization of sucrose molecules [29].

Thus, when plant molecules are extracted from plant matrices, we propose that these -OH groups, which are ionized, form a molecule-O-sucrose bond, which indeed aids in the solubilization of the plant molecules. Furthermore, below the pH (10–12) and temperature (30–60°C) levels, sucrose remains stable without getting hydrolyzed. The decrease in solubility of compounds over 50% sucrose may be due to saturation of sucrose molecules in water providing less space for compounds to solubilize. Also, it is observed from Figure 3 that pH 12 is sufficient to completely extract the compounds from plant matrices.

3.3 Effect of pH on precipitation

The next significant feat is to isolate the target molecule from other extracted molecules. As it is the pH level that stabilizes the solubilization of hydrophobic BDMS and UA into solution, neutralization could be tried to initiate selective precipitation. These studies and the corresponding observations are depicted in Figure 4. At pH 7, both BDMS and UA, being hydrophobic, completely precipitated from their respective extracts with significant purity, and the contaminating inorganic salts (like NaCl), which get thrown out usually in acid base extraction, were found to be still solubilized and posed no threat to the purity of the desired compounds extraction.

Figure 4:

Effect of pH (neutralization) on yield of precipitated β,β-dimethylacryl shikonin (BDMS) (left bar) and ursolic acid (UA) (right bar).

Both BDMS (c logP=4.21) and UA (c logP=8.63) are hydrophobic compounds that are solubilized into sucrose solutions due to alkalinity. Therefore, neutralization of alkaline extract selectively precipitates BDMS and UA leaving other compounds in solution. Also, the relative abundance of BDMS and UA in plant material establishes a concentration gradient, thus selectively precipitating BDMS and UA first from the aqueous alkaline sucrose extracts. This phenomenon was seen in precipitation of other molecules, too (e.g. Physcion, Rutin, Chrysin, and Naringin from their respective plant matrices).

3.4 Influence of extraction variables

The extraction time for the process was considerably reduced by the application of vacuum during extraction. Vacuum pulls the molecules out from the plant matrices, which under stirring works additively to lower the extraction time significantly from 48 h to 8 h (BDMS) and 56 h to 6 h (UA) (Supplemental Figure S7). Supplemental Figure S8 depicts the actual amount of vacuum and agitation required for complete extraction of compounds from plant matrices and their effect on extraction time. The observations reveal that vacuum at –10 (in Hg) and agitation (stirring) at 600 rpm are enough to attain maximum efficiency. The pattern also becomes stationary at higher values of vacuum and agitation.

In contrast to pH, extraction time, vacuum, and agitation, both temperature and solid loading (solid-liquid ratio) significantly affected the extraction efficiency. Supplemental Figure S9 shows that extraction becomes maximum at 50°C and 10% solid loading. Decrease of the extraction efficiency above these values might be because of de-stability of sucrose at higher temperatures and evaporation of water at higher temperatures under vacuum, which leads to increase in solid loading and concentration gradient limitations at higher solid loading ratios. The above observations were similar while extracting other compounds using the current process.

3.5 RSM: optimization

To study the applicability of the developed process in commercial/industrial practices that require scale up strategies, the process was modeled using RSM to acquire optimized parameters eventually leading to maximum yield of response variables (compounds). To validate the applied model to the process, it was applied to BDMS and UA, respectively, and the results were observed. From the above classical optimization investigations, it was clear that concentration of sucrose, temperature, and solid loading (three parameters) were found to influence the extraction process significantly, which were thus considered for RSM optimization. HPLC was used at each step of RSM to analyze the amount of metabolites. A mutual link between the dependent variables and independent parameters was established using the BBD and the second-order polynomial quadratic response equation (Eq. 1). It is represented by the equation below.

Table 1 depicts the results from extraction experiments and yield of compounds along with the Box-Behnken matrix. The statistical significance of quadratic model at each point was assessed by ANOVA. Tables 2 and 3 depict the ANOVA data demonstrating the statistical observations where the regression model has a high coefficient of determination (R²=0.998, 0.98). The significance of the model is explained by the R² adj (0.995, 0.95) values. There seems no significant difference between R² and R² adj values, which is desirable for the model. The low coefficient of variation (0.84, 3.71) indicated good reliability of process experiments. In the present study, Table 2 depicts that F values are greater than p values, implying that most of the coefficients obtained are significant in the model.

Table 3 depicts the coefficients and standard error. According to F values for coefficients, temperature (X₂) produces the major effect in extracting BDMS in the process (F value: 2048.26, p<0.0001) followed by solid loading (X₃) (F value: 72.96, p<0.0001). Concentration of sucrose (X₁) had the least effect. In contrast, for UA followed by temperature (X₂) (F value: 164.7, p<0.0001), concentration of sucrose (X₁) had next significant effect, whereas solid loading had least effect. Table 4 shows the results of the lack-of-fit test tabulating the variation in the data around the fitted model. In the present case, the F value for lack-of-fit test is 1.45, 1.1 (not significant), implying that the models sufficiently explain the obtained data. To evaluate the interaction among the operational factors along with finding the optimum values of each parameter, 3D response surface graphs were plotted based on obtained model equation. Figure 5 and Supplemental Figure S10 show the effects of influential parameters on yield of compounds. The yield of metabolites increases with increase in level of factor (0) and then decrease. For example, Figure 5 depicts the relation between concentration of alkaline sucrose (A) and temperature (B) where yield of metabolites increases as A increases from 40% to 50% and then decreases when it extends to 60%. Similarly, the yield reaches maximum as temperature (B) reaches 50°C and then decreases. This occurrence happened in all the surfaces drawn with interacting factors.

Figure 5:

Response surface plots for yield of β,β-dimethylacryl shikonin (BDMS) and ursolic acid (UA) showing the effect of operational parameters (X₁, concentration of sucrose; X₂, temperature).

Generally, numerical optimization method is preferred for optimization in which an input factor with desirable value and response can be selected. Using these conditions, the maximum achieved amount of BDMS was 18.2 mg/g DM at 50.91% of sucrose, 52.3°C and 9.47% of solid loading at –10 (in Hg) vacuum, and 1000 rpm agitation for 8 h. For UA, it was 2.2 mg/g DM at 53.81% of sucrose, 52.3°C and 9.93% of solid loading at –10 (in Hg) vacuum, and 1000 rpm agitation for 6 h. These results indicate an acceptable fit among the obtained data. It also agreed with the desirability of the model at all points. An added experiment was carried out to confirm the amount of compounds yielded at optimized conditions, which were 18.2 mg/g DM (BDMS) and 2.2 mg/g DM (UA). This was in accordance to the predicted values of 18.3 mg/g DM and 2.1mg/g (Table 5).

The final yield of 1.82% (BDMS) and 0.21% (UA) clearly supports a significant increase compared to initial yields of 0.72% and 0.07%, respectively, as given in Table 6. The final purity of BDMS and UA through optimized process was 99.3% and 96%, respectively, as analyzed by HPLC-DAD. The optimized yields signify an increase of 152% in total recovery of BDMS and 200% for UA, further establishing the potential for development and optimization of alkaline sucrose extraction procedure considering its impact on the process economics. Table 6 also depicts the complete observations obtained when the developed process alkaline sucrose precipitation was applied to separate other molecules from their respective plant matrices.

The fact that sucrose itself is a food additive (Table sugar) and the other solvents used in the current process makes it environmentally friendly, with minimum toxicity and minimum use of solvents, which considerably leads to a “green” route of production of commercially important metabolites as stated by Anastas and Warner [30]. The current process remarkably worked in separating different lead compounds from their respective plant matrices highlighting the wide applicability of a single process with minimum time required and less number of stages compared to current industrial manufacturing processes. However, more research is required for investigating the applicability of the process in separation of other important metabolites. The RSM modeling applied initially to BDMS and UA alone has to be investigated and applied to other metabolite separations, which will eventually lead to design an efficient scale up protocol for large-scale industrial application of the developed process.