Application of a Designed Mixed Model Approach for Antioxidant Extraction from Pomace

Pomace is a residue of juice production, but still holds many valuable compounds, such as antioxidants. However, efficient extraction proves to be challenging, as extractability is highly dependent on experimental conditions. Furthermore, antioxidants are also often retained by structural polysaccharides. Consequently, this study investigates the extractability of antioxidants in five different types of pomace in a 2 × 2 × 2 × 3 full factorial fully replicated design to determine the most efficient way of extraction. Therefore, extracting agent, temperature, extraction method and use of enzyme to break down structural components were alternated to investigate the effects of these parameters on the extractability. Main effects as well as interaction effects were estimated with linear mixed models. To study the effects of polysaccharides on the retention of antioxidants, the amount of soluble polysaccharides was measured and connected to the antioxidant capacity at each experimental condition. The experiments showed that acetone was most suitable as an extracting agent and that the use of pectinase significantly increased the amount of extracted antioxidants. Additionally, ultrasound-assisted extraction is advantageous compared to extraction in water bath. Temperature showed the lowest effect in this experimental design. Lastly, the results also highlight that the amount of soluble polysaccharides does not correlate with the extractability of antioxidants.


Introduction
Pomace is a residue, which describes the solid remains of fruits after pressing and mainly consists of skin, pulp, seeds and stems. Various studies have shown that pomace still holds a considerable amount of valuable nutrients, which is why the use of extraction could lead to an increase in the value chain (Perussello et al. 2017). Pomace could, for example, be used as a cheap source for pectin and/or antioxidants as food additives (Bruno et al. 2018;Morales-Contreras et al. 2020;Perussello et al. 2017).
Antioxidants are molecules, which inhibit reactions promoted by reactive oxygen species (ROS) and thereby reduce oxidative stress in human cells (Guo et al. 2020). Studies show that diminishing ROS can reduce the risk of several diseases, e.g. type 2 diabetes, cardiovascular diseases, cancer and chronic inflammations as well as influence the gut microbiome positively (Mithul Aravind et al. 2021;Skinner et al. 2018;Zhou et al. 2022). However, ROS also have important functions in the human body as second messengers, which balance radical production and scavenging in the cells. This is why a delicate equilibrium between ROS and antioxidants needs to be present in the cells (Clara et al. 2016;Guo et al. 2020). Typical antioxidants include polyphenols, vitamins, phenolic acids and nitrogen containing compounds. Various assays have been established to measure antioxidant capacity, one of them being the Folin-Ciocalteau assay (Folin and Ciocalteu 1927;Singleton et al.), which was developed for the determination of phenolic compounds; however, later was shown to also detect nonphenolic compounds, such as vitamins, thiols and nitrogen containing molecules (Everette et al. 2010;Prior et al. 2005). As this assay is sensitive to various compounds, it suffers from interreferences from, e.g. organic acids and sugars. It can however be assumed that within this study the interferences stay mostly the same.

3
The extraction of antioxidants from pomace, however, proves to be challenging. In general, most antioxidants are soluble, but some are bound to structural components. Consequently, various polysaccharide, e.g. pectin, prevent these substances from being extracted efficiently (Zhang et al. 2022). By degrading pectin using enzymes, the extraction yield of antioxidants from pomace can be improved as bound antioxidants also become available for extraction (Rani et al. 2021;Zhang et al. 2022). For improvement of understanding about the effectiveness of pectin degradation for antioxidant extraction, a variety of pomaces were chosen. Pomaces from fruits and vegetables with a high amount of structural components, e.g. apple and carrot as well as berry pomaces, were chosen to diversify the sample set. Furthermore, widely prevalent pomaces from, e.g. grape and apple, as well as minor pomaces from, e.g. chokeberry and currant, were chosen to further diversify the sample set. Additionally, efficient extraction of valuable compounds is often a difficult task, as the extracting agent, temperature and time as well as extraction method play a vital role in the extraction process (Egüés et al. 2021;Goldsmith et al. 2018;Meini et al. 2019). These parameters also affect various pomaces differently, which is why it is hard to estimate the true effect of enzymatic extraction in comparison with optimized extraction conditions without use of enzymes. Consequently, this study investigates the effect of various experimental conditions on the extractability of antioxidants for various types of pomace and compares them to the increase in extraction efficiency when pectinase is used to degrade polysaccharides. Additionally, the amount of pectin and its link to antioxidant extractability was investigated as well. Therefore, this study extensively investigates extractability of antioxidants from pomace in with the application of highly sophisticated mathematical models. This gives new and relevant insights into this topic.

Sample Preparation
Dried pomace from apple, chokeberry, black currant and grape was purchased from Holger Senger Vertrieb von Naturrohstoffen e.K. (Dransfeld, Germany). Dried pomace from carrot was made available by Kühner Ges.m.b.H & Co. KG (Peter am Hart, Austria). The samples were all already grinded; however, to ensure a large surface area, samples with greater particle size were grinded once more, which also achieved a more homogenous particle size distribution across all samples and consequently ensured similar extraction kinetics for all samples. Grape pomace had small but visible parts of wood in it, which were removed before grinding. Apple pomace also comprised some larger stem particles, which were removed as well. All samples were stored at room temperature until extraction and analysis.

Design of Experiment (DoE)
A 2 × 2 × 2 × 3 full factorial design with triple replicates was used to study the influence of temperature, use of enzyme, method of extraction and extracting agent on the extraction of antioxidants from dried pomaces of apple, chokeberry, carrot, currant and grape. The temperature was varied between 22 (RT) and 35 °C with the former serving as a reference. The latter temperature was chosen to ensure the preservation of all antioxidants (Galanakis et al. 2010). Additionally, extractions were performed with pectinase and without addition of enzyme, and the extraction was performed with water bath extraction (WB) under constant shaking and using ultrasonic extraction (US). The selection of extracting agents was based on a previous study (Srinivas et al. 2011) and comprised firstly ethanol extraction agent (EEA) consisting of a mixture of 80% ethanol, 19% deionized water and 1% formic acid; secondly, a methanol extraction agent (MEA), consisting of 80% methanol, 19% deionized water and 1% formic acid; and finally, acetone extraction agent (AEA) consisting of 80% acetone, 19% deionized water and 1% formic acid. Formic acid ensured the stability of certain antioxidants, mainly anthocyanins (Castañeda-Ovando et al. 2009). Each extract was produced under every possible combination of conditions three times.

Antioxidant Extraction
For antioxidant extraction, three different extraction agents were prepared: EEA, MEA and AEA (see the "Design of Experiment (DoE)" section). The extraction procedure was based on previously published methods (Lohani and Muthukumarappan 2016;Rana et al. 2015;Wiedemair et al. 2019) and adapted as necessary. All extractions were performed as triple replicates. Consequently, in total, 360 extractions were performed.
For antioxidant extraction without pectinase, 1 g of pomace was suspended in 15 mL of each extraction agent (EEA, MEA and AEA). The extractions were then, respectively, performed in a water bath under constant shaking for 30 min and an ultrasonic bath for 15 min. Additionally, all extraction were performed at 35 °C and at room temperature, which in consequently resulted in the 2 × 2 × 2 × 3 full factorial design described in the "Design of Experiment (DoE)" section. Then, the suspensions were centrifuged at 3700 rpm (3428 rcf) for 10 min, and the supernatants were collected in respective separate tubes. The residue was extracted under the same conditions two more times. All supernatants were combined, and the volume was adjusted to 45 mL with extraction agent. The extracts were stored in a freezer at − 20 °C until analysis.
For antioxidant extraction with pectinase, 5 mL of pectinase (85 u/mL) was added to 1 g of sample. The sample was mixed and left at room temperature for 30 min. After that, the same procedure as described before was used to extract antioxidants. Lastly, the volume was adjusted to 50 mL.

Antioxidant Capacity
Antioxidant capacity was determined photospectrometrically using the Folin-Ciocalteu assay (Wiedemair et al. 2019). Although this assay is often used to determine total phenolic content, many studies showed that it is also sensitive to other antioxidants (Everette et al. 2010;Prior et al. 2005). Therefore, 1.5 mL deionized water, 100 µL extract, 100 µL Folin-Coicalteu's phenol reagent (2N, Sigma-Aldrich) and 1.3 mL sodium carbonate (50 g/L) were mixed in a 4 mL PS cuvette (d = 1 cm) and put in an oven at 60 °C for 30 min. After cooling the room temperature and removal of air bubbles, the resulting blue coloured complex was measured at 750 nm with an UV/VIS spectrometer (Genova Plus, Jenway, Stone, Staffordshire, UK). Three repeat measurements were performed for each extract. Results are given in mg gallic acid equivalent (mgGAE) per g solid sample.
The standard calibration curve was recorded with each extraction agent (EEA, MEA and AEA) using gallic acid with 10 different concentrations between 0 and 600 mg/L. All concentrations were measured three times. The regression coefficient R 2 was 0.9993, 0.9986 and 0.9992 for EEA, MEA and AEA, respectively. Consequently, results are given as mg gallic acid equivalent (mgGAE) per g.

Extraction of Water Soluble Polysaccharides
Polysaccharides which are soluble in polar solutions, such as pectin and hemicellulose, were extracted by modified methods by Grana et al. and Wikiera et al. (Garna et al. 2007;Wikiera et al. 2016). Therefore, 1 g of dried pomace was suspended in 15 mL of 0.03 M HCl and heated to 90 °C in a water bath (Memmert GmbH & Co. KG, Schwabach, Germany) for 90 min. The suspension was then centrifuged at 3700 rpm (3428 rcf) for 30 min. The supernatants were filtered with a hydrophilic PVDF filter (0.22 µm), and an aliquot of 0.4 mL was then dispersed in 4 volumes of ethanol (1.6 mL). The suspension was centrifuged at 3700 rpm (3428 rcf) for 30 min and, afterwards, dried in a vacuum concentrator (Concentrator 5301, Eppendorf, Germany) at 45 °C to a constant weight. The precipitate was washed with 1 mL of 80% ethanol to remove sugars, pigments and other impurities and centrifuged again for 30 min at 3700 rpm (3428 rcf). The supernatant was discarded, and the pectin precipitate was again dried to constant weight in a vacuum concentrator at 45 °C. The amount of water soluble polysaccharides was then determined using an analytical scale.

Statistical Analysis
All statistical analysis and data visualization was performed using R (R Core Team 2021). First basic statistical parameters, such as means, medians and standard deviations, were calculated. Qualitative multivariate data analysis was performed using factor analysis of mixed data (FAMD) (Fac-toMineR package (Lê et al. 2008)) to gain a more detailed understanding of the data set. Then, a linear mixed model (LMM) with a Nelder-Mead optimiser (lme4 (Bates et al. 2015)) was computed to estimate the influence of the various conditions and their interaction on the measured amount of antioxidants. The model was fit by restricted maximum likelihood (REML). Analysis of variances (ANOVA) was used to estimate the significance of all factors and interactions. To investigate, whether or not all pomaces have a similar response to varying experimental conditions, additional LMMs were computed for each type of used and compared pomace.
For the estimation of the influence of the application of pectinase specifically, relative increases for each experimental pair of parameters were computed for each type of pomace, and the effects were modelled in a linear model (LM) with QR factorization method as well.

Antioxidant Capacity
The amount of extracted antioxidants varied greatly depending on type of pomace. Figure 1 shows the mean antioxidant capacities of all performed experiments. The highest amount of antioxidants was extracted from grape pomace and the lowest from carrot pomace.
In general, the antioxidant capacity varied between 1.04 and 33.33 mgGAE/g with a mean of 12.39 mgGAE/g and a median of 10.98 mgGAE/g. Ideally, mean and median would be almost identical, which is not the case for this data set due to the stark differences in antioxidant capacities of various types of pomace. The standard deviation across the whole data set was 9.52 mgGAE/g. Antioxidant capacity for apple, currant, carrot, chokeberry and grape pomace across all extraction conditions varied from 2.29 to 7.23 mgGAE/g, 4.18-20.10 mgGAE/g, 1.04-3.44 mgGAE/g, 10.26-32.30 mgGAE/g and 15.73-33.33 mgGAE/g, respectively. All of these results are in accordance with previously published literature (Antoniolli et al. 2015;Dulf et al. 2018;Iora et al. 2015;Jabbar et al. 2015;Kapasakalidis et al. 2009;Lohani and Muthukumarappan 2016;Rana et al. 2015;Sánchez-Rangel et al. 2021;Sójka and Król 2009). The antioxidant capacity of currant was influenced the most by change of extraction conditions, whereas the antioxidant capacity of grape was influenced the least.
The mean antioxidant capacities of apple, currant, carrot, chokeberry and grape pomaces across experimental conditions were 4.15 mgGAE/g, 11.01 mgGAE/g, 2.28 mgGAE/g, 21.05 mgGAE/g and 23.45 mgGAE/g, respectively. The medians were 3.64 mgGAE/g, 11.27 mgGAE/g, 2.05 mgGAE/g, 19.78 mgGAE/g and 23.57 mgGAE/g, respectively. These means and median are more similar because they reflect a more homogenous subset of the whole data set. Figure 1 as well as the reported means clearly shows that carrot pomace had the least amount of antioxidants, followed by apple, currant, chokeberry and grape. Additionally, an increase of antioxidant capacity is visible after treatment with pectinase for all types of pomace. Antioxidant extraction was most efficient using AEA, whereas treatment with EEA and MEA yielded similar results. The use of acetone is favourable for extraction of larger antioxidants, such as flavonols (Shah et al. 2014), which are highly present in fruits. Ethanol and methanol both exhibit similar polarity and target the extraction of small antioxidants. In previous studies, methanol has often shown to be more efficient for antioxidant extraction than ethanol; however, its toxicity is a concern (Shah et al. 2014). Figure 1 also shows that differences between US and WB extraction seem to vanish if higher temperatures are applied or pectinase is used to make antioxidants more readily available. At lower temperatures, ultrasonic extraction is more efficient. It is known that US is favourable for the extraction of thermolabile components. Furthermore the vibration by ultrasound ensure a constant mechanical mixing between sample and solvent, which is

Multivariate Data Analysis and Linear Mixed Models
To get a more detailed qualitative overview over the influences on the extraction process, a factor analysis for mixed data (FAMD), which can be interpreted similarly to a principal component analysis (PCA), was computed (see Fig. 2). Figure 2 shows that the first dimension of FAMD, which explains almost 20% of the variance in the data set, is mostly influenced by the type of pomace. Consequently, this parameter influences the amount of antioxidant capacity the most. This is also the reason why carrot and apple as well as grape and chokeberry have a similar position on the x-axis -they hold similar respective amounts of antioxidants. Additionally, EEA and MEA are separate from AEA, which suggests that EEA and MEA act similarly in extraction, but different from AEA. The use of enzyme is positioned at diagonally between dimension 1 and 2, which suggests that is contributes equally to both. The categories of extraction method and temperature are located near the coordinate origin, which indicates that they only explain little variance in the data set.
To further investigate the various effects of different extraction conditions quantitatively, statistical analysis using a mixed model approach over all pomaces' and experimental conditions was computed. Therefore, use of enzyme, extracting agent, extraction method and temperature were set as fixed effects and type of pomace as well as the each condition nested in type of pomace was set as random effect. Significance levels of main effects and interactions were estimated using type 3 ANOVA. The pseudo coefficient of determination (R 2 ) was computed with generalized variance approach. The results are shown in Table 1.
Regarding main effects, all but temperature are significant. This confirms the qualitative analysis from Fig. 2, where temperature contributes to neither the first nor second dimension of the FAMD. The estimates also show that the use of AEA as extracting agent has the largest effect on antioxidant extraction as it has a higher significance (p < 0.01). Overall extraction conditions, the use of MEA for extraction has a significant advantage over EEA; however, the p-value is close to 0.05. Use of pectinase as well as implementation of an US extraction method also has a significant influence  Table 1, all but one (Extracting agent (EEA):Extraction method (US)) two-way interactions are significant (p < 0.05). The same is true for three-way interactions with only Use of enzyme (w/o enzyme):Temperature (RT):Extraction method (US) being non-significant. All four-way interactions were shown to be significant.
The pseudo coefficient of determination for linear models can be used to interpret the amount of variance explains by main effects as well as interaction. According to Table 1, the largest amount of variance in the model stem from the parameters use of enzyme as well as extracting agent. R 2 also confirms that temperature is the only main effect, which only explains little variance. The coefficient of determination also highlights that interactions do not contribute to explaining variance in the model, which means that most likely the interactions between parameters are only significant because their constituting main effects are. The highest R 2 for interactions is 0.104 for Temperature (RT):Extraction method (US), which is also similar to the R 2 of the main effect for type of extraction (US vs. WB).
The interactions were further investigated. In Fig. 3, the modelled predicted values for interactions regarding antioxidant capacity between all levels are depicted. The error bars correspond to the standard error stemming from the different types of pomaces and are in correspondence with the statistical evaluation of random effects in the linear mixed model. There, type of pomace was revealed to have the highest variance (88.7415), followed by extracting agent (5.5807) and use of enzyme (3.1645). Figure 3 shows that in general the influence of extraction method becomes less pronounced at higher temperatures. Furthermore, the effects of temperature and extraction method become less pronounced when applying pectinase.  ( Additionally, the antioxidant capacities yielded with EEA and MEA are quite similar when using pectinase, regardless of temperature, but not without the use of pectinase. If no enzyme is applied, methanolic extracts yield a higher antioxidant capacity compared to ethanolic extracts. The highest amounts of antioxidants occur in AEA extracts. Regarding AEA, the effect of temperature and use of pectinase seems to have a high impact on antioxidant capacity according to Fig. 3. Figure 1 and 2 further suggest that varying experimental conditions impact the types of pomace differently. This is why mixed linear models were also computed for each type of pomace separately. In Table 2, the results regarding the significance of main effects and interactions are presented. Regarding main effects, all but temperature and extracting agent (EEA) are significant for every type of pomace. The influence of temperature is not significant for apple pomace and only slightly significant for grape pomace. The effect of extracting agent (EEA) is not significant for carrot pomace. Looking at interactions, it becomes clear that less interactions are significant for specific pomaces than for the overall LMM. The only two-way interaction significant for all types of pomace is Temperature:Extraction method. Regarding three-way interactions, no interactions is significant for every type of pomace, and Temperature:Extracting agent (EEA):Extraction method is non-significant for every investigated type of pomace, except apple. Regarding fourway interactions, no interaction is significant for all types of pomace, however Use of enzyme:Temperature:Extracting agent (AEA):Extraction method was significant for all but apple pomace. Use of enzyme:Temperature:Extracting agent (EEA):Extraction method was only significant for carrot pomace. This more detailed analysis clearly shows that each pomace reacts differently to changing experimental conditions. However, the overall assessment that most main effects have an influence and still remain.

Amount of Water-Soluble Polysaccharides and Relative Increase of Antioxidant Capacity
In addition to the overall effects of extraction conditions, the influence of the amount of pectin and other soluble polysaccharides on antioxidant extraction potential was also investigated. Therefore, water-soluble polysaccharides were extracted and quantified from all pomaces in triplicates. Then, the relative increases in extracted amount of antioxidants before and after pectinase, treatment was also calculated. Therefore, the change in antioxidant capacity was analysed for each sample at all possible conditions of a specific variable (AEA, EEA, MEA, WB, US, 35 °C, RT). The The amount of water-soluble polysaccharides was determined to be 16.6%, 6.2%, 11.6%, 6.7% and 13.5% for apple, currant, carrot, chokeberry and grape pomace, respectively. Then, the total increase and relative increases for each experimental parameter after application of pectinase were calculated and are shown in Fig. 4. Figure 4 depicts that the influence of use of pectinase on the extractability of antioxidants is dependent on the extraction conditions. Regarding relative increases, the minimal relative increase was 4.41% and occurred when pectinase was applied to chokeberry, which was extracted using MEA with US extraction methods at 35 °C. The maximal relative increase was 200.58% and occurred when carrot pomace was extracted with AEA using WB extraction at room temperature. The mean and median relative increase over all experimental conditions was 62.67% and 48.48%, respectively.
The mean and median vary from each other, which indicates that the relative increases are not equally distributed across experimental conditions, which is confirmed by Fig. 4. The illustration shows that when extraction takes place at room temperature, the increase is in extracted antioxidants is higher than at 35 °C. Additionally, the increase is also more pronounced when using water bath extraction. The effect of the application of pectinase is also dependent on the extraction agent. In methanolic extracts, the increase is the least pronounced, whereas in AEA and EEA are similar in their increase of extracted antioxidants after enzyme addition.
To further investigations, the influence of relative increases after application of pectinase, a LM was computed. The results showed that the extracting agent (p = 9.55•10 −05 ) as well as the temperature (p = 0.04201) significantly influence the relative increase of extracted antioxidants. Consequently, while differences between relative increases in extracted antioxidants for different extractions methods (WB Table 2 Results for the mixed model analysis by restricted maximum likelihood (REML) for each pomace. AEA acetone extracting agent, EEA ethanol *Significant effect (p < 0.05), **Significant effect (p < 0.01), ***Significant effect (p < 0.001), n.s. not significant
The addition of pectinase also influences the investigated types of pomaces differently. The highest increases are found for currant and carrot pomace. While carrot has many structural components, which are most likely deconstructed with pectinase, currant is fairly a soft berry. This already hints that the amount of pectin in the fruit does not correlate well with the increase in extracted antioxidants. This was confirmed by all coefficients of determinations being below 0.2. The lowest increases regarding antioxidant capacities occur in chokeberry and grape across all conditions.
In general, the relative increase for apple, currant, carrot, chokeberry and grape pomace ranged from 19.33 to 185. 08%, 40.72-199.70%, 17.52-200.58%, 4.41-74.62% and 11.12-85.37%, respectively. The means were 69.15%, 98.56%, 79.12%, 32.78% and 33.73%, and the medians are 55.93%, 78.23%, 68.28%, 24.62% and 32.00%, respectively. Previously published literature also showed that antioxidant extraction is increased after application of enzyme (Chamorro et al. 2012;Costoya et al. 2010;Kitrytė et al. 2017;Landbo and Meyer 2001;Meyer et al. 1998;Will et al. 2000). For apple pomace, Will et al. reported an increase between 60 and 80% after enzyme application (Will et al. 2000). For currant pomace, increases between 12 and 65% were reported for different conditions (Landbo and Meyer 2001). Kitryte et al. reported an increase of approximately 30% in extractability of antioxidants after enzyme application in chokeberry pomace (Kitrytė et al. 2017). Antioxidant capacity from grape pomace could be increased between 19 and 42% by Chamorro et al. (Chamorro et al. 2012). Relative increases of 13-28% were reported by Khandare et al. for black carrot juice (Khandare et al. 2011). Consequently, the reported relative increases in literature after application of enzyme also show that the amount of structural polysaccharides is not necessarily correlated with the relative increase of extracted antioxidants. This is also supported by the data of this study.
The reason for this may be that, for example, for grape and chokeberry pomace, most antioxidants are already available without degradation of structural elements, which is supported by low relative increases. Consequently, depending on the type of pomace, polysaccharides may not hinder the extraction of antioxidants.

Conclusion
The presented study investigated antioxidant extraction from 5 different types of pomaces in detail using a 2 × 2 × 2 × 3 full factorial design with triple replicates. US extraction was found to be advantageous over water bath extraction, especially considering room temperature extractions. Furthermore, AEA yielded higher antioxidant capacities as EEA and MEA, and lastly, pectinase significantly increased antioxidant capacities of extracts. These results were supported by FAMD analysis. Interactions between main effects were also shown to be significant, although these variables only contribute little to explain variance in the data set. Consequently, it is likely that interactions are only significant because their constituting main factors are. Fig. 4 Relative increases in antioxidant capacities after application of pectinase at various experimental conditions. AEA, acetone extracting agent; EEA, ethanol extracting agent; MEA, methanol extracting agent; RT, room temperature; WB, water bath extraction; US, ultrasonic bath extraction No significant correlation between the amount of pectin in the respective pomaces and relative increase in extracted antioxidants could be established. Consequently, the application of enzyme to increase extractability of bioactive compounds has to be tested for each type of pomace, in order to assess the benefit of enzyme treatment, which is also supported by previously published literature. In general, enzyme treatment was most effective for currant and carrot pomace and least effective for grape and chokeberry pomace. For the latter two, this means that antioxidants are mostly not bound to polysaccharides and could thus already be extracted effectively before.
Overall, our experiments proved that each type of pomace reacts differently to changing experimental conditions. Consequently, a detailed analysis and a suitability study of all results are necessary to uncover valuable information also for the industrial scale-up.