Comparisons of quantification of total polyphenols using LC-DAD single wavelength and MCR-ALS

In this study, a comparison of the total amounts of different polyphenol classes (phenolic acids, flavonoids, and anthocyanins) were compared by employing LC-DAD and applying MCR-ALS signal processing. The LC-DAD strategy involves quantification at a single wavelength selected to represent the target compounds, while in MCR-ALS, the whole spectra of the compound are considered after unmixing from other components. However, for quantification, both methods employ a single reference standard. Total phenolics (usually quantified at 280 nm) encompass a diverse group, including phenolic acids, flavonoids anthocyanins, and other compounds. Spectral overlap among these groups complicates estimation, particularly distinguishing between polyphenols like phenolic acids and anthocyanins. To appreciate the significance of the overlaps between compounds of different classes, their molar absorptivity is compared at different wavelengths as shown in Table S2 (see ESM). Compounds with high molar absorptivity such as anthocyanins and flavonoids give a considerable signal at 280 nm; therefore, estimating phenolic acids is almost impossible with single wavelength quantification. Several other interferences such as amino acids and organic acids have been shown to increase the risk of overestimating the phenolic amounts as they also have significant absorptivity at 280 nm (Table S2 in ESM). Therefore, at this wavelength, selectivity is generally lacking. However, after applying the MCR-ALS algorithm, the total phenolic acids were able to be estimated without the influence of anthocyanins, flavonoids, and some small organic acids and amino acids, compared with the amounts estimated by LC-DAD single wavelength as shown in Fig. 2. However, since the identities of the compounds were mainly based on spectral profile, it has to be pointed that it is still uncertain that the amount estimated by MCR-ALS could be all phenolic acids, as other amino acids might exhibit same spectral profile as target compounds. However, the contributions of some small organic acids such as maleic acid, ascorbic acid, and amino acids such as tryptophan were confirmed by their reference standards after MCR-ALS extraction (Fig. S2, see ESM). The initially high estimation of flavonoid contents before MCR-ALS also suggests contributions from compounds other than flavonoids (Fig. 2), which could likely be phenolic acids absorbing at slightly higher wavelengths such as chlorogenic acid, as well as anthocyanins. However, total anthocyanin shows no significant difference between quantification with and without MCR-ALS processing, confirming the selectivity of the high wavelength (520 nm) used (Fig. 2).

Fig. 2

Total phenolics (GAE), total flavonoids (QE), and total anthocyanin (CE) content were extracted by four different extraction procedures and quantified using two different strategies (LC-DAD-single λ and MCR-ALS). Total phenolic compounds extracted by MCR-ALS exclude flavonoids, anthocyanins, and other interfering organic acids. FW represents fresh weight, while DS and FS refer to dried (freeze-dried) samples and fresh samples, respectively. Error bars represent the standard deviation for n = 3 extractions

More importantly, the significance of this observation varies among extraction methods; for instance, flavonoids extracted by the PLE technique show minimal difference before and after MCR-ALS, indicating fewer interfering components. Conversely, after MCR-ALS, other methods exhibit approximately half the amount of flavonoids recorded by LC-DAD single λ quantification, suggesting more interfering compounds. Interestingly, UAE assisted with acidified methanol as a solvent also indicated a high difference between the total phenolic amounts estimated by LC-DAD-single λ and MCR-ALS compared with other techniques, which might suggest that the method is more comprehensive in extracting several compounds. While this could be reflecting the differences in extractability of components (target compounds and interferences) between extraction conditions of techniques employed, it could also possibly suggest that the nature of interferences varies between these extraction methods. In addition, some variation might be due to the difference between the freeze-dried sample and the fresh samples.

Comparisons of extractability of total polyphenols between extraction methods

In general, investigating the extractability of three different groups of phenolic compounds (phenolic acids, anthocyanins, and flavonoids), employing the four different extraction protocols indicated that UAE-FS yielded the highest extractability (P < 0.05), regardless of the quantification strategy employed (Fig. 2). This increased extractability with MeOH as extraction solvent compared with PLE where the solvent is mostly water may stem from its lower solvent polarity, which facilitate the extraction of a wide range of phenolic compounds. Similar studies have supported the efficacy of water-alcohol mixtures in particular, as the best solvent for flavonoids and anthocyanins compared with water (only) infusions, as demonstrated in lingonberries [32, 33]. Extraction in acidic conditions proved highly favorable for anthocyanins, stabilizing their cation-charged flavylium form [34], as well as improving peak shapes in HPLC analysis. Sample preparation by centrifugation of the aqueous layer from the fresh sample, while advantageous for preserving thermally labile phenolic compounds due to its lack of heating, resulted in the poorest extraction of polyphenols, possibly due to its inability to extract polyphenols bound to the matrix. Conversely, PLE using water/ethanol/FA (95/4/1, v/v/v), under pressurized conditions, exhibited lower extractability of the phenolics, flavonoids, and anthocyanins compared with the acidified methanol with UAE. The elevated temperatures in PLE (100 °C) may decrease the viscosity, aiding solvent penetration of the matrix to access bound phenolic compounds, yet it could also lead to compound degradation. In contrast, milder temperatures (40 °C) employed in UAE may be suitable for balancing extractability and minimizing degradation.

The observed discrepancies could not only be attributed to the effect of solvent mixtures but also other factors such as extraction time, pressure, and temperature which were different between techniques. The difference between fresh and freeze-dried samples also marks a major contribution in the observed variation extractability (compared with the UAE method), where higher extractability was observed in fresh samples compared to freeze-dried samples, among all compounds groups (Fig. 2). To further explore the cause of these difference, an investigation was conducted by adding of the same amount of water content as fresh sample (85.5% moisture) to the freeze-dried samples (Fig. S3, see ESM). The results revealed about 30% increase in anthocyanin content from the re-moisturized freeze-dried sample, compared with the dry berry smoothie sample (Fig. S3, see ESM). However, the addition of water did not alter the extractability of total flavonoids. Nonetheless, it is not possible to conclude if these observations are an indication of losses freeze-during drying or alteration of the sample morphology due to freeze-drying affecting the extractability of compounds. However, loss of anthocyanins during the freeze-drying process has been previously reported [35], linking them to increased enzymatic activities resulting from cell disruption and prolonged freeze-drying time.

Comparisons of selectivity of different extraction methods

In this study, data sets from different extraction protocols were also subjected to MCR-ALS to evaluate and compare various extraction protocols regarding the significance of interferences relative to target polyphenolic compounds, as indicated in the flow chart in Fig. 1. As already indicated, quantification typically conducted at a single wavelength may be hindered by spectral overlaps from interfering compounds. The precision of quantification also relies on the extraction method’s selectivity, with highly selective methods producing cleaner extracts with fewer spectral overlaps. MCR-ALS was used to estimate the nature of interferents among different extraction methods, given its ability to provide information about the interferent’s spectra. Interferences were evaluated in three chromatographic ranges (3–20, 20–45, and 45–55 min), yielding satisfactory fittings for all extraction methods, with explained variances greater than 99% and lack of fit less than 10%. As shown in Fig. 3, MCR-ALS has shown the ability to enhance chromatographic resolution, which allowed for the investigation of overlapping unknown interfering peaks. More importantly, interfering peaks were observed in the region between 3 and 20 min where most phenolic acids such as gallic acid eluted and 45–55 min where most flavonoids eluted. Overall, the region where anthocyanin eluted (20–45 min), showed very minor interferences. In the 3–20 min chromatogram region, major interfering compounds exhibited absorption spectra around 248 nm (Fig. S2, see ESM), which are likely organic acids, particularly intense in the extract from PLE and centrifugation.

Fig. 3

Chromatograms (top) and corresponding spectral profiles (bottom) showing target and overlapping interferences obtained after MCR-ALS analysis of different elution ranges, i.e., 3–20 min (top, left), 20–45 min (middle), and 45–55 min (top, right)

Relative selectivity values, ranging from 0 to 1 (where 1 represents maximum selectivity), were calculated by comparing confirmed target compounds with interfering peaks (Fig. 4). Results indicated that UAE with acidified MeOH as a solvent exhibited better selectivity on average compared with other extraction protocols, particularly for early eluting compounds such as phenolic acids and anthocyanins (Fig. 4). For instance, for the observed peaks eluting between 3–20 min, we can estimate that 72% of the total observed signal intensity can be attributed to phenolic compounds, while 28% accounts for other interferences when using UAE. Conversely, the PLE technique and centrifugation of the fresh sample showed more overlapping non-target peaks, likely due to water contributing to the extractability of polar compounds like amino acids, an observation in line with the previous study [31]. However, this also suggests that these methods enable the high extractability of these small untargeted organic acids, compared with UAE with methanol. Limited selectivity for flavonoids eluting in the 40–45 min region was observed across all extraction techniques, likely due to the ability to extract non-polar organic compounds like terpenes. It should be stated that some of the major overlapping compounds in the region where flavonoids eluted were phenolic acids such as p-coumaric acid, which were not considered as interferences for the calculation of extraction selectivity since they are part of the target compounds when considering the whole spectrum. Although elucidation of all possible interfering peaks might be a challenging task and a limitation in this study, a screening with high-resolution mass spectrometry and MS DIAL library search helped to annotate both target compounds (Table S1, see ESM) and few interferents (Table S3, see ESM). The details of mass spectrometry experiments are shared in S1. As shown in Fig. S4 in the ESM, the potential interferents were plotted in a scatter plot to visualize how they are retained relative to different groups of analytes of interest (phenolic acids, flavonoids, and anthocyanins) on a C18 column. It is quite clear that most of the early eluting peaks are small organic acids such as ascorbic acid, quinic acid, and maleic acid. In addition, some of these interferences were further confirmed by the use of standards in HPLC, and the peaks are marked in Fig. S2 in the ESM. These organic acids are quite common in berry fruits [36, 37]. Other interferences eluting in the region where most of the target analytes eluted were possibly various heterocyclic compounds and terpenes (Fig. S4, see ESM). While these interferences might not all have a significant impact on the quantification of anthocyanins and flavonoids as they absorb light at fairly high wavelengths, their contribution to the analysis of other compounds like phenolic acids could be impactful, as shown by the molar absorptivity data in Table S1 in the ESM.

Fig. 4

Relative selectivity of different extraction methods on the extraction of phenolic compounds (excluding anthocyanins and flavonoids), anthocyanins and flavonoids, against other interfering compounds as elucidated by MCR-ALS. Error bars represent the standard deviation for n = 3 extractions. DS refers to freeze-dried samples, while FS is fresh samples

Evaluating the effect of overlapping peaks on single compound quantification

Further characterization was carried out by identification and quantification of individual peaks forming the phenolics, flavonoids, and anthocyanins (annotated in Fig. S1, see ESM). The method of choice employed for extraction was UAE (fresh sample) with acidified MeOH, as extraction solvent, as it exhibited the highest extractability and fair selectivity among all the compound groups. It can be noted that the majority of peaks detected from 26 to 60 min were anthocyanins which also show a signal at 280 nm (Fig S1, see ESM). Nineteen compounds were confirmed by their retention times, UV spectra, and mass spectra in the berry smoothie samples, of which 7 compounds belong to the anthocyanin group, 7 flavonoids, 7 phenolic acid and 1 stilbene (Table S2, see ESM). Of these detected compounds, 15 compounds were quantified relative to their reference standards as shown in Table 1. It was observed that 3-O glycosylated anthocyanins were detected in higher amounts followed by chlorogenic acid. Several findings reported cyanidin, delphinidin, pelargonidin, peonidin, malvidin, and petunidin, in their glycosylated form, as the most anthocyanidins distributed in the various berries [38, 39]. Phenolic acids content were generally in low amounts (some even below the limit of quantification), and only chlorogenic acid was found in substantial amounts. Chlorogenic acid has been also identified as the main phenolic acid in blueberries [40].

Table 1 Amounts of single compounds after quantification with both LC-DAD and LC-DAD with MCR-ALS data processing, in berry smoothie sample extracted with UAE using 1% FA in methanol. Paired t-test was used to compare the significance of the variability of quantification between the two methods employed (p-value)

However, to establish how much the overlapping interferences affected the quantification of the detected compounds, the quantification of single compounds using both HPLC-DAD and MCR-ALS processing was compared (Table 1). Since the true concentration in the sample is unknown, differences between the methods may indicate impurities or variability due to the quantification methods used. However, we can assume that an agreement between the two methods confirms the purity of the peak as well as the quality of the chromatographic separation, given the ability of MCR-ALS to detect impurities covered under the peaks. Overall, most peaks indicated valid purity, showing p-values above 0.05 when comparing quantification with LC-DAD and MCR-ALS. This was particularly evident for the anthocyanins, where high selectivity is expected even with single wavelength quantification. Interestingly, flavonoids, which exhibited numerous interferences among all the extraction techniques, also showed less difference between individual quantified peaks with both LC-DAD and MCR-ALS, except for quercetin 3-O glucoside which indicated slightly higher variability. Despite several interferences observed in the region where flavonoids eluted, closer evaluation indicates that most of the interferences did not co-elute with the identified target compounds. These also approve the contribution of chromatographic resolution considering that these compounds eluted late (45–60 min) allowing for better resolution due to increased retention. In addition, other interferences showed very low absorptivity at 360 nm, rendering the quantification at that wavelength selective.

Evaluating the long-term stability of polyphenols using LC-DAD and EFA

The developed method was utilized to study the degradation of polyphenols quantified in Table 1. However, the stability study was focused on flavonoids and anthocyanins only, as the number of phenolic acids identified was limited and their concentrations were low. The berry smoothie samples were extracted with UAE with 1% FA in MeOH. LC-DAD was utilized for quantification, while MCR-ALS was the supporting tool used to survey interfering peaks. EFA, in particular, was utilized to elucidate the degradation products, without the use of reference standards. The findings were also used to model the degradation kinetics of polyphenols following storage at various temperatures.

In general, anthocyanins showed the most variability as a function of temperatures compared with flavonoids. Most anthocyanins maintained stable concentration for almost 18 and 12 months in the freezer (Fig. 5) and fridge (Fig. S5, see ESM), respectively, mostly showing insignificance differences in concentration over the studied time (p < 0.05). In fact, the degradation of anthocyanins and flavonoids in berry smoothies stored in the fridge and freezer did not fit according to the first-order kinetic model (Eq. 3), except for cyanidin 3-O glucoside (Fig. S6, in ESM), which showed an average half-life of 28 months for both fridge and freezer storage (Table S4, see ESM). However, the degradation of anthocyanins in berry smoothies at room temperature followed the first-order kinetic model (Fig. 6), with individual anthocyanins exhibiting half-lives ranging from 7 to 12 months (Table S4, ESM). The results of the fitting coefficients such as rate constant, coefficient of determination, and root mean square error are shown in Table S4 in the ESM. The anthocyanins storage at room temperature also showed similar degradation rates as reported by Muche et al. [41], for anthocyanins studied in grapes. However, fridge and freezer storage slowed down anthocyanin degradation, approximately four times better than room-temperature storage. Low temperatures reduce the activities of enzymes in the plant cells, particularly polyphenol oxidases and peroxidases responsible for phenolic degradation [42, 43]. Moreover, it is also reported that temperature may influence the equilibrium state of anthocyanins, favoring the chalcone form [44]. Flavonoid contents, as well as the pH of the berry smoothie, were among the variables that remained stable for at least 12 months period across all storage temperatures (Figure S7–10, see ESM). However, a notable drop was observed after 9 months of room temperature and fridge storage for some of the glycosylated flavonoids, probably degrading to their aglycone forms.

Fig. 5

Comparison of anthocyanins response over 30 months in berry smoothie stored in freezer (− 20 °C). Error bars represent the standard deviations for n = 3 samples extracted

Fig. 6

Comparison of anthocyanins response over 12 months of berry smoothie stored at room temperature (fitted with first-order kinetic model). Error bars represent the standard deviations for duplicate samples

The data from room temperature storage samples, where most degradation was observed, was subjected to EFA, to elucidate potential degradation products. From this data set, the singular value decomposition computed five eigenvectors as the representation of the variation in the data. However, only three values were associated with chemical meaning, which was attributed to anthocyanins, other phenolic compounds, and intermediates formed during the process, while the other values were associated with noise in the data. The resultant model led to a lack of fit and explained variance of 0.96 and 99.99%, respectively. The results of EFA are shown in Fig. 7, where the forward EFA plot and the backward EFA plot are overlaid. The scale on the y-axis represents the log EV (eigenvalues) of each PCA analysis versus the storage time (x-axis). Therefore, the lines connecting the homologous eigenvalues were found to depict the spectra of anthocyanins (shown by forward EFA) and the formation compounds likely to be phenolic derivatives (shown by backward EFA). This also attests to the reports that suggest the degradation of anthocyanins through the hydrolysis of the glycosidic bond connecting the aglycone and glycosyl group, ultimately leading to the formation of products like aldehydes and benzoic acid derivatives [45, 46].

Fig. 7

Eigenvector plot derived from EFA, where the orange line shows the forward EFA and the blue line shows the backward EFA, spectral corresponding to anthocyanins and other phenolic derivatives, respectively

The trend obtained with EFA corresponds well with the degradation of anthocyanins as determined by LC-DAD. Clearly, the break-even point between the degradations and formation of phenolics occurs at around 6 months, which is similar to the average half-life of anthocyanins predicted by first-order kinetics. EFA has been applied previously to predict the kinetic model of bioactive compounds exposed to various conditions [47]. In our study, we correlated the EFA prediction in parallel with utilizing reference standards. By performing EFA on augmented chromatograms from only baseline to 1 months of storage, new peaks of compounds, suspected to be organic acids, were observed, with lambda max around 248 nm (Fig S11, see ESM). These could also suggest that the other compounds break down into small organic acids during room temperature storage of the berry smoothie. The results obtained with EFA were also confirmed by inspecting MCR-ALS profiles from deconvoluted chromatographic data for each storage temperature. The finding also suggests spectral profiles and peaks of some phenolic compounds, found primarily in samples stored for 6–12 months at room temperature (Fig. S12, see ESM), indicating the formation of these compounds after degradation. These outcomes provide insights into the evolution of species during these kinetic processes, especially where no standards are available.