Bilberry Press Cake
Press cake was acquired from a commercial bilberry juice producer in northern Sweden. A randomly chosen unpretreated press cake of 8 kg was freeze dried (Edwards E2M5 − 50 °C and 8 hPa) and milled (Retsch Mixer mill MM 400) to produce a homogeneous powder that was kept frozen until use.
The three factors used for the experimental design were total solids (TS), mass % of ethanol used, and temperature. Temperature and % ethanol are expected to influence the extraction. The amount of extracted polyphenols should increase as the TS goes up but it was important to check this factor for nonlinearities. The extraction experimental design for methods 1 and 2 was based on a slightly modified three-factor Box-Behnken (BB) design with 12 runs (Myers and Montgomery 2002). To this design, 6 star points were added and the center point was replicated 3 times, resulting in 21 runs in total. A central composite circumscribed (CCC) design with 12 runs was used for the two-phase experiments (methods 3 and 4).
The TSE used double the solvent volume of the RBR so the TS levels were also doubled. Extraction time is also a factor but it was not included in the models. Instead, each run was checked over time: TSE for 0, 8, 16, 32, and 64 min and RBR for 2, 4, 8, 16, 32, and 64 min. This way each of the times has its own complete design available. Times were chosen for practical reasons. RBR allowed measurement after 2 and 4 min while for TSE, this could not be done.
The levels used for ethanol were in mass %: star low = 10, low = 20, center = 40, high = 60, and star high = 70. For temperature (°C), the star points were chosen inside the BB design for practical reasons in star low = 30, low = 20, center = 40, high = 60, and star high = 50. For the RBR-dried press cake (g) used for extraction, the levels were star low = 1.05, low = 3.33, center = 5, high = 10, and star high = 12.27 while for the classical extraction; these masses were 2.11 g, 6.66 g, 10 g, 20 g, and 24.5 g (approximately double those for RBR). This is a deviation from the ideal textbook values, but still possible to use for calculations.
Extraction of Anthocyanidins
Traditional Solvent Extraction (Method 1)
The solvent mass used was 400 g each time, with ethanol amounts of 40 g, 80 g, 160 g, 240 g, and 280 g and the remaining part to a total weight of 400 g was filled with distilled water. The different design temperatures were maintained by using a climate chamber. Magnetic stirring at 300 rpm was used. After temperature equilibration in the climate chamber, the press cake powder was added. Ten-milliliter samples were taken out and filtered after 0, 8, 16, 32, and 64 min; these were used for FC and NIR measurement. After 64 min, two 15-mL samples were taken out and sent for HPLC analysis (Fig. 2).
Rotating Bed Reactor (Method 2)
In the RBR reactor (SpinChem), the solid is added to and contained in a rotating drum and the solvent circulates through a chamber due to the centrifugal forces created by the rotation. Figure 3 gives a schematic of the flows and the working principle of the RBR reactor. The RBR can be temperature controlled by pumping water from a thermostatic bath though the jacket around the chamber. All pumping was done with peristaltic pumps. Sampling was done after 2, 4, 8, 16, 32, and 64 min for NIR and FC measurement. Final samples taken after 64 min were analyzed by HPLC.
Two-Phase Extraction (Method 3 and 4)
A special technique tested in the present paper is two-phase extraction where salt solution/ethanol mixtures generate a two-phase system when high concentrations of certain salts are added. The two-phase system has an additional advantage that polyphenols end up in the ethanol phase while sugars end up in the water phase which did not display color. With many other extraction techniques, the sugars remain mixed with the anthocyanidins.
The system of two-phase extraction was tested with two different salts namely NaH2PO4 and (NH4)2SO4 (respectively methods 3 and 4). For each salt, an experimental design in the factors TS and temperature was carried out. The extractions were carried out over time using the intervals 2, 4, 8, 16, 32, and 64 min.
Method 3 used: 96 g H2O, 54 g Na2HPO4 × 2 H2O (VWR, Analar), and 50 g EtOH.
Method 4 used: 102 g H2O, 38 g (NH4)2SO4 (VWR Analar), and 60 g EtOH.
To summarize, the compared extraction methods were as follows:
Traditional extraction using different concentrations of ethanol dissolved in water, different masses of material, and different temperatures.
RBR extraction using different concentrations of ethanol dissolved in water, different masses, and temperatures.
RBR extraction using two-phase system, method 3 only mass and temperature were changed.
RBR extraction using two-phase system method 4, only mass and temperature were changed.
Analysis of Total Phenolics and Anthocyanidins
Total Phenolic Content
The Folin-Ciocalteu (FC) reagent (Sigma Aldrich) is a mixture of tungstenate/molybdate/phosphoric acid reagent in hydrochloric acid that acts as a colorimetric assay for polyphenolic antioxidants, proteins, and some other analytes. The recipe used was one optimized for polyphenols. The intensity of the blue color is related to the amount of polyphenols present in solution. The blue color was measured in a transmission cuvette at 765 nm (Singleton et al. 1999). A calibration curve was made using gallic acid (Sigma). Transmission measurement was done using a Shimadzu UV-1800 (Shimadzu Corp, Japan).
Visible and NIR spectroscopy can be used to identify and quantify many colored and other organic chemicals (Burns and Ciurczak 2001; Siesler et al. 2007). The extract was measured by a Tec5 VIS-NIR spectrometer (Tec5, Germany) using a homebuilt transmission flow cell with 5-mm transmission path length. The liquid was pumped through the flow cell. The Tec5 gives responses at wavelengths 305–2200 nm, but the ranges above 1400 nm were omitted because it mainly represented noise and irrelevant information, see Fig. 2. The Agrospec software was used to control the Tec5.
Anthocyanidin Measurement by HPLC
The content of individual anthocyanidins, delphinidin, cyanidin, petunidin, pelargonidin, peonidin, and malvidin, was measured by HPLC in the literature (Swadseh 1996). Our instrument was a Shimadzu LC-20A Prominence equipped with an autosampler. Eluent used for analysis was (A) 0.4% TFA in water and (B) 0.4% TFA in acetonitrile. Gradient program was as follows: 0–4 min, 84% A; 4–15 min, 84–62% A; 15–16 min 62% A; 16–16.1 min 84% A; 16.1–17 min 84% A, stop. Analytical separation of anthocyanidins was carried out using a Poroshell 120 SB-C18 (4.6 × 75 mm; 2.7 μm) column. The column oven was set at 30 °C and the flow of the mobile phase was 1.0 mL/min. Anthocyanidins were detected at 530 nm using a photodiode array 190–800 nm.
All the standards (delphinidin, cyanidin, petunidin, pelargonidin, peonidin, and malvidin chloride) were purchased from Extrasynthese. Water was ultrapure quality water (Millipore) and acetonitrile was of HPLC grade. Stock solution for each individual standard was 5 mg/mL. Working standards were diluted as follows: 62.5, 125, 250, and 500 μg/mL where each standard contains all six individual anthocyanidins. Standards were used to generate a four-point calibration curve. Identification was based on retention time and UV-spectra of the single standard.
Multivariate Data Analysis
Much emphasis in the paper is on data analysis, both analyzing the experimental designs and analyzing the multivariate spectroscopic data. All the experiments were carried out according to experimental designs. For analysis of a design, multivariate linear regression (MLR) is used for modeling responses from experimental settings and their transformed values (Box and Draper 1987). The designs were subjected to a complete analysis by regression coefficients and determination of their significance. Response surface figures were made for some designs. The experimental designs were analyzed in the Modde 10.1 software (Umetrics, Sartorius Stedim Biotech, Sweden). For the spectroscopic analysis by VIS-NIR, multivariate analysis by PCA and PLS was used. Evince software (Prediktera, Sweden) was used for this purpose.