Food-grade NaHCO₃ was purchased from local supermarket (Florence, Italy). Other reagents and solvents (analytical or HPLC grade) were acquired from Merck (Saint Louis, MO, USA). α+β-Punicalagin (purity ≥ 91%) was acquired from Phytolab (Vestenbergsgreuth, Germany); GA (purity ≥ 99%) and EA (purity ≥ 95%) were acquired from Extrasynthese S.A. (Lyon, Nord-Genay, France).

Pomegranate (Punica granatum L.) peel samples of Wonderful and G1 varieties were purchased from Azienda Agricola Onori Maria Rosaria (Fermo, Marche, Italy) and Rio del Sol soc. Agricola (Faenza, Emilia Romagna, Italy). Peel samples were air-dried at 42°C for 3 days, cut into raw pieces, and packed into zip-lock bag until extraction trials. Before extraction, the dried peel samples were crunched using a blender (IKA^® M20 Universal Mill, Staufen Germany) and filtered through a 60-mesh sieve to obtain a fine and homogeneous powder. The peel samples and their extracts used are listed in Table 1. Preliminary tests to determine the best hydrolytic conditions were performed on G20M variety samples; the chosen methodological parameters were then applied to treat four peel samples.

Conventional extraction, called decoction, was used for comparison with hydrolyzed extracts, using a published method (Khatib et al., 2017) with some modifications. Briefly, 2.5 g of dried peel were dissolved in 50-mL ultrapure water and heated at 95°C for 1 h with continuous stirring, using AREX-6 Digital PRO (Velp Scientifica SRL, Usmate, Italy) heating-magnetic stirring plate. Pomegranate tannins are non-thermolabile at the decoction temperature (Khatib et al., 2017). The solution was centrifugated (Z323 K high-speed centrifuge with cooling system, Hermle Labor Technik GmbH, Germany) for 10 min at 5,000 rpm and room temperature; the supernatant was recovered and used for HPLC-DAD-MS analysis.

Acid hydrolysis was performed directly on the peel without any previous extraction by applying a published method (García-Villalba et al., 2015), as follows: peel powder (0.05 g) was placed in 5 mL of 4-M HCl. The sample was vortexed for 1 min and placed in oven at 90°C for 1, 4, 8, and 24 h; then, it was cooled down at room temperature, adjusting pH to 2.5–3.5 with 4-M NaOH, values suitable for the chromatographic column used for HPLC analysis. The formed precipitate (pellet) was removed by centrifugation (10 min, 3,500 rpm), and the supernatant was analyzed by HPLC-DAD-MS.

The pellets were washed, testing two types of solvents: (i) 10-mL dimethyl sulfoxide–methyl alcohol (DMSO–MeOH), 50:50 (v/v) as proposed previously (García-Villalba et al., 2015); (ii) 10-mL ethyl alcohol–water (EtOH–H₂O), 80:20 (v/v). After dissolution, the sample was vortexed for 2 min, then centrifuged (5,000 rpm, 10 min) and the supernatant was analyzed by HPLC-DAD-MS.

Bicarbonate-assisted extraction is a laboratory-developed method conducted as described: peel powder (2.5 g) was added to 50 mL of either 0.3% (w/v) or 0.6% (w/v) NaHCO₃ aqueous solution; the two concentrations were chosen after preliminary extraction tests using concentration of NaHCO₃ ranging 0.1–3.3%. The sample was maintained under stirring on the heating plate for 60 or 120 min at an extraction temperature of 95°C. It was then centrifuged (5,000 rpm for 5 min) and supernatant was recovered. Only for HPLC analysis, the supernatant was cooled down and acidified to pH 2.5–3.5 with HCOOH, causing formation of a precipitate, which was separated after a second centrifugation (5,000 rpm, 10 min), thus obtaining second supernatant (i.e., the “extract” sample) and a precipitate. The latter precipitate was washed with 25 mL of 80:20 (v/v) EtOH–H₂O solution to recover co-precipitating phenolic fraction; the obtained mixture was sonicated for 5 min in an ultrasonic bath (Argolab, Carpi, Italy) and centrifuged (5,000 rpm, 5 min), thus obtaining a third supernatant (i.e., the “washing” sample). “Extract” and “washing” samples were collected and analyzed through HPLC-DAD-MS.

Tests on bound phenols were conducted on sample W20 M variety (Table 1) with a laboratory-developed method based on total extraction of free phenols by decoction plus different washes, and successive hydrolysis of the exhausted peels to verify the release of free phenols from polymeric matrix. The decoction was done with 2.5 g of peels in water (50 mL) at 95°C for 60 min (Khatib et al., 2017); then, the residual solid was recovered after centrifugation and washed twice with 25-mL EtOH–H₂O, 80:20 (v/v) in ultrasonic bath (5 min) to recover residual free phenols and the unknown polar compound (UPC). The washed solid residue was then extracted with either HCl or NaHCO₃. AH was performed with 50-mL 4-M HCl (4 h, 90°C) and then pH adjusted to 2.5–3 with 4-M NaOH. The hydrolysis with NaHCO₃ was performed with 50-mL 0.6% NaHCO₃ solution (1 h, 95°C). Both solutions were centrifuged at 5,000 rpm for 10 min, the supernatants were recovered, and the extracts were analyzed by HPLC-DAD-MS. All extractions were performed in triplicate.

The yields of extracts from NaHCO₃ hydrolysis (chosen as the best method to obtain lower molecular weight phenols-enriched extracts) were determined as dry extract weight; 10 mL of extract (decoction or hydrolysates with NaHCO₃) were freeze-dried using Leybold Heraeus LYOVAC GT2, GEA Lyophil GmbH, Düsseldorf – Germany. For the extracts with NaHCO₃, the net weight of dry extracts was calculated subtracting the estimated salt content from the total extracts’ net weight. Tests were carried out in triplicate.

HPLC-DAD analysis was performed using an Agilent HP1200 liquid chromatograph equipped with a DAD detector (Agilent Technologies, Palo Alto, CA, USA) and a core-shell Kinetex column C18 (100 × 3 mm, 2.6 μm; Phenomenex, USA) by following a published method (Cecchi et al., 2023) with little modifications. Mobile phases were: (a) water/formic acid (pH 3.2), and (b) CH₃CN. The following multi-step linear solvent gradient was used: 0.1–8 min, 5–25% B (v/v); 8–18 min, 25% B; 18–20 min, 25–95% B; 20–26 min, 95% B; 26–28 min, 95–0% B; and 28–32 min, 0–5% B. Elution time was 32 min, with a 10-min post-time and a flow rate 0.4 mL/min. The ultraviolet-visible (UV–Vis) spectra ranged from 200 nm to 500 nm with chromatograms acquired at 210, 254, 280, and 370 nm. The MS experiments were performed using an Agilent HP 1260L liquid chromatograph equipped with a HP1260 (G6125B) Mass Spectrometer Detector with an API/electrospray interface (Agilent Technologies). The electrospray ionization (ESI) parameters were: nitrogen flow rate of 10.5 L/min, drying gas temperature of 350°C; nebulizer pressure of 1,811 Torr; and capillary voltage of 3,500 V. Acquisition was performed in full spectrum scan (range 100–2,000 Th). The experiments were carried out in a negative ionization mode by applying fragmentation voltages of 150 V and 180 V.

For quantitation of phenolic compounds in extracts, three calibration curves were built using standard solutions of α+β-punicalagin, EA, and GA, as follows:

Gallic acid: λ = 280 nm, five data points, linearity range 0–2.5 µg; R² = 0.9979. This curve was used to quantify GA and UPC from AH, since the compound presented a similar UV spectra of GA with peak at 280 nm.

Results were expressed in mg/g of dry matter (DM) (i.e., dried peel or dried extract [DE]).

Each experiment was performed in triplicate and results were expressed as mean ± SD using the EXCEL software (version 2013) in-house routines. Statistical significance of the quantitative data was assessed with one-way ANOVA and F-test at p < 0.05, which were performed using the EXCEL software (version 2013). The least significant difference (LSD) procedure was used as a post hoc comparison (DSAASTAT software v.1.1, Onofri, Pisa, 2007). Two-way ANOVA was carried out with the Origin Software, version 2024b (Northampton, MA, USA) to assess significance at p < 0.05.

The AH of pomegranate peel samples with hydrochloric acid was used for comparative purposes by applying a previous analytical protocol (García-Villalba et al., 2015). The type and concentration of hydrochloric acid and the process temperature were the same as already reported (4-M HCl at 90°C), while variations in extraction time and washing method of the solid residue after extraction were applied. Figure 1 shows the main changes that occurred from 1 to 24 h for the G20M sample selected to perform a preliminary screening focused on choosing the best extraction time.

The first difference from the reference study (García-Villalba et al., 2015) was the presence of an unknown polar compound (UPC, Figure 1A) in all samples, not detected previously. The molecule showed a retention time of only 2.95 min on a reverse phase column and a maximum absorption at 280 nm; it achieved maximum amount at 8 h and was not found either in decoction or after extraction with bicarbonate. UPC was not recognized as a derivative of the hydrolysis of pomegranate tannins because its chemical features did not confirm the presence of a phenolic structure (see Supplementary Material). It did not respond to electrospray ionization either in negative or positive mode, not even applying different fragmentation energies, and the proton spectrum did not show signals attributable to phenolic rings. Further studies are underway to define the structure of UPC and its precursors.

The phenolic profiles of G20M sample at different times of AH are reported in Figure 1B, along with the decoction profile as a reference. Punicalagin concentrations presented a strong decrease after 1 h of hydrolysis and disappeared at longer extraction times. Simultaneously, the concentration of punicalin increased significantly, compared to decoction after 1 h of hydrolysis because punicalagin lost hexahydroxydiphenoyl (HHDP) moiety. For 1 to 4 h, punicalins were stable, likely because of an equilibrium between the continuous release determined by the hydrolysis of precursors and their further degradation; after a prolonged time of hydrolysis (8–24 h), both punicalagin and punicalin disappeared. These data are partially in contrast with those reported previously by García-Villalba et al. (2015), which found punicalins up to 12 h of acid extraction. Compared to decoction, EA was significantly increased in 1-h acid hydrolysate but decreased at longer hydrolysis times because of low solubility in acidic water (Bala et al., 2006).

Formation of a precipitate and its composition are discussed in Section 3.3. After the screening of G20M sample, extraction of 4 h was applied on Wonderful samples to investigate any phenolic variations at the time of punicalin degradation. Extraction time of 8 h was also applied because of the major recovery of interesting compounds in washing samples (Table S1; Section 3.3), allowing observations of differences among water extracts and the insoluble phenolic fraction. Figure 2 compares the kinetics of phenolic compounds in the water extracts obtained from the Wonderful peel samples after 4- and 8-h hydrolysis with HCl.

The trend observed for phenols in sample G20M (Table S1) was also confirmed for other Wonderful samples. In addition, in this case, UPC was produced, reaching its maximum concentration after 8 h.

Punicalin isobar was one of the main phenols in aqueous extracts and it increased in all samples after 4 h (Figure 2); a different behavior among peel samples was observed at 8 h, because only for W21ER and W22ER samples, it increased further, but a slight decrease was observed for other samples. Such different behaviors are attributable to differences in peel composition. Concerning α- and β-punicalins, all samples showed the same trend with a maximum increase after 4 h, while prolonged reaction time determined almost complete degradation. In agreement with the data shown in Figure 1B and the poor solubility of EA in acidic water, it partially precipitated determining low concentration in extracts and a high variability among replicates, not observed for other molecules. Total compounds, which as a sum include all minor phenols detected in the extracts, resulted lower if compared to the decoction (Figure 2F), mainly because lipophilic molecules, such as EA and GA released by the degradation of punicalagins and punicalins, partially precipitate in the acidic environment. A slight increase for some samples after 8 h of hydrolysis was observed because UPC was also considered. Overall, the acidic extraction proposed previously (García-Villalba et al., 2015) to hydrolyze the tannins of pomegranate showed several critical issues, such as a strong degradation of native phenols and use of HCl as a chemical agent, thus appearing unsuitable for future scale-up processes applied to pomegranate peels.

Concerning extraction of pomegranate peel with bicarbonate solutions, preliminary tests with variable NaHCO₃ concentrations (0.1–3.3% w/v) and reaction time (20–120 min) were performed to select suitable salt concentrations and process time. A complete degradation of tannins occurred when exceeding 1% (w/v) of bicarbonate with a reaction time of 60 min or more, and similar results were obtained with a reaction time of 20 min and a bicarbonate concentration of 3.3% (data not shown). Along with the reaction time of 60 and 120 min, lower concentrations of bicarbonate (0.3% and 0.6%) were successively selected to reduce final salt content in the extract. Further tests were conducted with G20M sample within the defined experimental ranges. The distribution of main molecules after hydrolysis with 0.3% w/v NaHCO₃ (B1) and 0.6% w/v NaHCO₃ (B2) for 60 and 120 min, compared with decoction (D), are reported in Figure 3.

For both concentrations, a partial hydrolysis of original tannins was observed with a higher degree of hydrolysis for B2 samples. A higher degradation of β-punicalagin and α-punicalin was reached in B2 samples (B2, 60 min). Furthermore, EA showed the same concentration in both B1 and B2 samples after 60 min. Overall, significant decreases of α- and β-punicalagin were observed in all samples, compared to decoction. Considering the results in Figure 3, the extraction time of 60 min and the bicarbonate concentrations of 0.3% and 0.6% w/v were selected to treat all peel samples listed in Table 1.

As reported in Figure 4, GA and EA were the main hydrolysis products in all samples, their content increased proportionally with the increase of bicarbonate, with a similar trend for α- and β-punicalin.

Differently, α- and β-punicalagin, the main precursors of previous molecules, decreased proportionally with the increase of bicarbonate. It is worth highlighting that the presence of punicalagin isobars (compounds 12 and 13 in Table 3) which increased with the same kinetics of EA, results in major compounds in bicarbonate extracts. Their origin could be hydrolysis and rearrangement of one ester bond of precursor molecules (α- or β-punicalagin). Overall, the data shown in Figure 4 highlighted that a partial hydrolysis of native tannins was obtained in non-alkaline environment, because the sample solution from NaHCO₃ at 0.3% was associated with pH values close to neutrality, suggesting a possible catalytic effect of bicarbonate in the hydrolysis of punicalagin’s ester bonds, not associated with an alkaline environment. Finally, among the extracts of the same peel sample, the total phenolic amount of the supernatant, which included all minor phenols, did not show significant variations from decoction to B1 and B2 samples (Figure 4F). In light of the data in Figure 4, extraction with 0.6% bicarbonate for 1 h could be proposed as a simple method to increase the amount of low molecular weight phenols, including EA and to reduce the concentration of punicalagins, without reducing total amounts. Further discussion on total phenols is in presented Sections 3.4 and 3.5.

Since part of the released phenols during hydrolysis are lipophilic molecules, with low water solubility, a washing step was applied to solid residues formed after AH and for the precipitates formed after the acidification of bicarbonate extracts before the HPLC analysis of supernatant. In particular for EA, its planar structure confers a high degree of crystallinity and a poor solubility in water with a maximum of 9.7 µg/mL at neutral or acidic pH (Bala et al., 2006). To accurately estimate EA and other lipophilic phenols released after hydrolysis, the washing samples were evaluated by HPLC-DAD (results presented in Tables 2A, 2B, and S1).

Concerning acid extraction, two washing mixtures were tested: DMSO–MeOH, 1:1 v/v, as an ideal solvent for good recovery of water-insoluble phenols as indicated in the literature (García-Villalba et al., 2015), although it hinders the drying process because of high boiling point and is not of food-grade nor applicable for possible scale-up processes; EtOH 80% v/v was selected because it was of food grade, certainly usable for a scale-up, and easily removable, thus enabling the acquiring of dry extracts.

The effectiveness of two different solvents in recovering insoluble phenols from the solid residues of G20M and W20M samples was evaluated preliminarily (Table S1). The washing samples were characterized by the presence of valoneic acid dilactone, gallagic acid dilactone, and EA. Regarding recovered phenols, DMSO–MeOH solution was more exhaustive than EtOH 80% solution, giving approximately a double amount of total phenols, with EA being by far the most abundant compound. Hydroalcoholic solvent was, however, suitable for recovering a good part of precipitated phenols and was therefore applied to other peel samples of the Wonderful variety: two major compounds significantly changed during acidic hydrolysis, namely valoneic acid and EA, while total phenols differed among samples at different times, and significantly increased at 8 h in all washing samples (Table 2A).

Concerning the extraction with bicarbonate, the formation of a precipitate was only observed at the end of extractive steps, during the pH adjustment with formic acid required for HPLC analysis. Consequently, such precipitate cannot be considered a residue of the one-step extraction but a fraction of the whole aqueous extract from bicarbonate. Indeed, the 80% EtOH-wash samples showed similar profiles as those observed for the corresponding extract (supernatant) with only increased amounts of EA, especially in the samples extracted with 0.6% NaHCO₃ (Table 2B). The significance of independent variables (sample and time for AH and sample and bicarbonate concentrations for hydrolysis with alkaline agent) and their interaction according to 2-way ANOVA are reported in Tables S2A and S2B, respectively. The residue produced after AH (Table S2A) contained mostly valoneic acid dilactone, GA, and EA. However, time significantly affects only valoneic acid and GA. Interestingly, no overall effect of time was observed on EA content, although the interaction between sample and time was significant (p < 0.05). Nevertheless, the total content was significantly affected (p < 0.001) by both time and sample type.

Regarding the residues obtained after hydrolysis with bicarbonate (Table S2B), the levels of GA and EA were significantly affected by bicarbonate concentrations. For the native α-punicalagin and β-punicalagin, both compounds were significantly affected by sample, bicarbonate concentration, and their interaction. Similarly, the amount of new isobar of punicalagin was significantly influenced by both sample type, bicarbonate concentration, and their interaction (p < 0.001).

Total phenols extracted from dried peels were evaluated in the extracts of Wonderful peel samples, because this is one of the most important and widely cultivated varieties globally. Differently, the G1 variety was included in this study for different morphology of the mesocarp (much thinner than the Wonderful variety), although it is cultivated only in Italy. From the data discussed above, it was observed that this different morphology is not related to relevant variations in the mesocarp’s tannin profile.

Total phenols extracted were calculated as sum of phenols recovered in both supernatant and EtOH 80% washing samples (Figure 5).

Phenols after AH were statistically comparable to those of the decoction for 8-h hydrolysates, while in the case of 4-h hydrolysates, amounts were significantly lower for all samples. B1 methodology (60 min with 0.3% bicarbonate) was the one that extracted maximum phenols for all samples, significantly higher than in all other extracts. A small decrease in B2 samples was observed due to the stronger punicalagin degradation when a greater quantity of bicarbonate was used, according to data shown in Figures 3 and 4. However, the B2 method still recovered slightly higher amounts, compared to decoction when the residue was considered.

Considering the high content of phenols extracted through bicarbonate and the possibility to use these samples for future biological tests also due to the increased content of low-molecular weight phenols, including EA, the percentage yields of the dried extracts of dry peels and TPC (mg of phenols/g of DE) were determined.

Because the bicarbonate extraction protocol required the addition of sodium salt and subsequent acidification with formic acid prior to HPLC analysis, sodium formate was present in liquid extract estimated at 2.1 mg/mL and 4.3 mg/mL for 0.3% and 0.6% bicarbonate, respectively. The corresponding amounts were subtracted from the total weight of each dry extract for obtaining net percentage yields of dried peel (Figure 6A).

For each sample, the mean net yields increased proportionally to the bicarbonate concentration, reaching a maximum of about 61% for G20M variety sample in the case of 0.6% of bicarbonate. The yields obtained from decoction were always lower, ranging 43–53%, with changes depending on the origin of peel sample. It was supposed that the use of bicarbonate facilitated the release of further polar compounds from the peel, compared to decoction.

The TPC was consequently evaluated as mg/g of DE (Figure 6B). The highest TPC was found in the B1 extracts of all samples with a range of 241.0–361.3 mg/g DE. In B2 samples, TPC was always significantly lower than decoction and B1 samples (172.9–285.9 mg/g DE), although with slight differences. The differences in tannin content among samples collected over three different years (Figure 6B) resulted from significant variations in levels of tannin expressed on dry peel and were consistent with the data shown in Figure 4F. Variations in tannin richness among pomegranate peel samples, even within the same variety, is attributed to geographic origin, climatic conditions, and age of tree, with older plants producing lower levels of phenolic compounds (Dadáková et al., 2020). Within each peel sample, the use of bicarbonate allowed producing final dry extracts with increased or comparable amounts of TPC, compared to a reference sample, such as decoction, but otherwise characterized by higher amounts of low-molecular weight phenols.

Hydrolytic procedures, such as acid or alkaline hydrolysis, are often applied to evaluate the amount of bound phenols in different vegetal or fruits (Zhang et al., 2020). The presence of bound phenols in pomegranate peels is scantly investigated in the literature (Dadwal et al., 2017; García-Villalba et al., 2015; Sun et al., 2021). The authors of these studies, although applying different extraction procedures, observed an increase in phenolic content after hydrolysis, suggesting that a part of phenols in pomegranate peel is covalently bound to polymeric structures (e.g., through ester bonds to cellulose and pectin).

In this study, to prove the presence of bound phenols in pomegranate peel and to determine their content, decoction for 60 min was selected as a conventional extraction. The peels of W20M variety sample were first extracted by decoction (D), then the solid residue was submitted to two washing steps (Wash 1 and Wash 2) to remove any residual phenols, and finally the solid residue was hydrolyzed with HCl (acid hydrolysis) or bicarbonate at 0.6% w/v (B2). As shown in Table S3, Wash 1 contained approximately 10% of phenols extracted in hot water and only 1.2% of phenols were in Wash 2 sample, while after acidic hydrolysis of the washed residue, only UPC was detected (data not shown) and no other phenolic compound was revealed. The extract with 0.6% of NaHCO₃ contained 3.7% of total phenols (around 6 mg/g of DM), with 2.8 mg/g of EA and limited amounts of punicalins and punicalagins (from 0.3 to 1.3 mg/g DM). Based only on alkaline hydrolysis, the presence of bound phenols in pomegranate peels was confirmed, but in very low amount.

Bound phenols in other fruits (e.g., cranberries, strawberries, red grapes, grapefruits, apples, pineapples, pears, bananas, peaches, lemons, and oranges) showed a similar content, ranging 2–24% of total phenols in common fruit (Sun et al., 2002). Additionally, the measured phenols released after mild alkaline hydrolysis included mainly ellagitannin derivatives, known to be little water soluble, leading to possible underestimation. Literature studies estimated the content in bound phenol on pomegranate peels of 50% (Dadwal et al., 2017). Nevertheless, in the latter case, because the measurements were done by a colorimetric assay, an overestimation of real content was hypothesized, as already observed for chestnut tannins passing from HPLC to Folin–Ciocalteu results (Khatib et al., 2023). A different study that discussed bound phenols in pomegranate peels found few isomers of punicalagins along with valoneic acid; however, no quantitative evaluation was carried out. In this case, the main detected compounds were glycosides, sterols, and pentacyclic triterpenoids compounds (Sun et al., 2021), whose recovery was mostly driven by the non-polar nature of the solvent used for suspending the exhausted residue of peels.

Bound phenols or non-extractable phenols were estimated in 11 pomegranate peel samples (without specifying varieties), assessing that after acidic hydrolysis it was possible to recover up to 36% more phenols (non-extractable fraction) than by conventional extraction (García-Villalba et al., 2015). In our study, only 3.7% of non-extractable phenols were found in the Wonderful variety (Table S4) by not applying the acid extraction but the extraction with bicarbonate 0.6%. Differently from our study, the study conducted by García-Villalba et al. (2015) used conventional method for extractable (or free) ellagitannins by a single-step extraction at room temperature with 0.1% of HCl in MeOH/DMSO/H₂O (4:4:2, v/v/v), with 10 min of extraction time. No other extraction or washing was successively applied to solid residue to verify whether the extraction was exhaustive.

Such a large difference is mainly related to different extraction procedures applied to evaluate free or extractable tannins. In this research, different reasons were indicated to use non-acidified hot water as the best extraction mixture to recover free ellagitannins: (i) ellagitannins are well-soluble in this medium and stable at 100°C; (ii) the hot treatment is suitable to better disaggregate the structure of peel and favors the extractability of phenols; and (iii) hot extraction for 60 min was proven to be exhaustive (Khatib et al., 2017).

Although our data revealed a very low amount of bound phenols in pomegranate peel from the Wonderful variety, further research, including a larger number of samples, is needed to confirm this result for the peels of other less studied pomegranate varieties.

The complete list of phenolic compounds detected in both hydrolyzed and non-hydrolyzed extracts is given in Table 3, which shows presence/absence of the compound in decoction (reference method without hydrolysis) and extracts from hydrolysis methods. The compounds were tentatively identified by the UV-Vis spectra, retention time, comparison with data from the literature (Akande et al., 2022; Gu et al., 2013; Hemingway and Hills, 1971; Hernández-Corroto et al., 2019; Man et al., 2022; Plumb et al., 2002; Sentandreu et al., 2013; Tanaka et al., 1990; Tuominen and Sundman, 2013) and by the MS fragmentation pattern in negative ionization mode.

Compound 2 with molecular ion at mass-to-charge ratio (m/z) of 781 was identified as an isobar of α-punicalin and β-punicalin with a lower retention time. This hypothesis is explained by the possible rearrangement of a punicalin intramolecular bond to reach a more stable structure in acidic environment. In fact, this isobar was detected in acidic hydrolysates after longer extraction time (8 and 24 h) where precursor punicalins were completely degraded.

Compounds 3 and 4 were tentatively identified as HHDP-glucose isomers, mainly because their mass spectra showed molecular ions at 481 m/z, typical of this structure. These compounds were detected in both decoction and extracts hydrolyzed with bicarbonate.

Compounds 5 and 6 were identified as GA monohexoside and GA, respectively, because of the presence of molecular ion at 331 m/z for glycoside, and at 169 m/z for GA. Identification of GA was confirmed by comparison with pure standard. Small amounts of GA were detected in decoction, and its concentration increased proportionally to the bicarbonate concentration or to the reaction time during AH (see Sections 3.1 and 3.2). GA increase during hydrolysis is explained by the degradation of many minor compounds containing galloyl residues, such as compounds 20, 30, 33, etc.

Compound 7 was tentatively identified as galloyl-HHDP-hexoside according to the molecular ion at 633 m/z (Figure 7). The molecule was detected in both decoction and extract hydrolyzed with 0.3% NaHCO₃ but was absent in 0.6% NaHCO₃ extract and in the samples obtained by AH. Also, compound 31 (rt = 7.52 min) presented the same molecular ion and similar UV-VIS spectrum, allowing its identification as an isomeric form of galloyl-HHDP-hexoside. This isomer was detected in all extracts. Applying extract ion function at 633 m/z, it was possible to find several isomers, which presented this ion in their mass spectrum as a molecular peak; however, since their concentration was barely detectable, only two isomers are shown in Table 3.

Compounds 9 and 10 were identified as α- and β-anomeric forms of 4,6-gallagyl-glucoside (punicalin), as they showed [M-H]^- fragment at 781 m/z and [M-2H]^2- ion at 390 m/z. Formation of double charged ions is frequently observed in MS negative ionization mode, especially when tannins are in high concentration in the sample, and these adducts confirm the molecular weight of analyte (Akande et al., 2022).

Compound 11, detected as a main compound only in acid hydrolyzed extracts, has not been identified so far. Several attempts were made to identify the chemical nature of this molecule (HPLC-DAD and solid phase extraction [SPE] purification, followed by HPLC-DAD-MS, ESI-ionic trap analysis, matrix-assisted laser desorption ionization (MALDI) analysis, and proton nuclear magnetic resonance [¹H-NMR] analysis), described in detail in Supplementary Material. The molecule did not respond to electrospray ionization by applying different fragmentation voltages in negative or positive mode. Furthermore, data from the (¹H-NMR) spectroscopy experiments showed the absence of aromatic protons in its molecular structure (see Section S1.4 in Supplementary Material).

Compounds 12 and 13, detected in small amounts in decoction but in higher quantities in the extracts with bicarbonate, were classified as reaction products after rearrangements of intramolecular bonds of punicalagin. They were identified as punicalagin isobars with a molecular fragment at 1,083 m/z and a double charged ion at 541 m/z. Compounds 15 and 23 were referred to as α- and β-form of punicalagin, according to their spectral data and the retention time of standards.

Compound 16, detected in all analyzed extracts, was identified as valoneic acid dilactone in agreement with molecular ion at 469 m/z and the ion derived from decarboxylation at 425 m/z. The same fragmentation pattern was observed for compound 40, identified as an isomer of valoneic acid dilactone, probably sanguisorbic acid dilactone (García-Villalba et al., 2015).

Compound 18, showing the main fragment at 483 m/z, was recognized as digalloylhexoside. This compound was detected in the decoction and in bicarbonate-hydrolyzed samples, but not in acid-hydrolyzed extracts.

Compounds 19 and 22, with an ion at 1,039 m/z and the corresponding double charged ion at 519 m/z, were identified as isomeric forms derived from the decarboxylation of punicalagin; the molecules were found for the first time only in pomegranate peel extracts from bicarbonate.

Compounds 20 and 33, both with molecular ions at 785 m/z, were identified as isomeric forms of digalloyl-HHDP-hexoside, and together with compound 30 (935 m/z) identified as galloyl-bis-HHDP-hexoside, and were detected in decoction but not in hydrolysates.

Compound 21 was recognized as digalloyl-HHDP-gluconic acid according to the fragment at 801 m/z, attributable to molecular ion. This compound was identified in decoction and in bicarbonate-hydrolyzed samples but not in AH extracts.

Compound 24 was tentatively identified as brevifolin carboxylic acid showing molecular ion at 291 m/z and the ion derived from decarboxylation at 247 m/z; the structure is reported in Figure 7. This EA derivative previously reported in Tanaka et al. (1990) is formed starting from DHHDP group.

Compound 26, with a molecular ion at 799 m/z, was tentatively identified as lagerstannin A, previously found in pomegranate juice (Sentandreu et al., 2013).

Compounds 27 and 29 were detected only in acidic extracts and identified as pedunculagin isobars due to a molecular ion at 783 m/z and UV-VIS spectra, very similar to those obtained for pedunculagins. Compounds 32 and 34 were detected in all extracts and identified as α- and β-pedunculagin. It was hypothesized that compounds 27 and 29 are formed after a rearrangement of ester bonds of pedunculagin structure, as observed for punicalins and punicalagins.

Compound 28 was tentatively identified as gallocatechin pentoside, showing a corresponding UV spectrum and a molecular ion with MW of 437 m/z. Presence of this phenolic derivative at low concentration in pomegranate peels was first reported by Plumb et al. (2002).

Compounds 36 and 38 were identified as α- and β-anomeric forms of the same ellagitannins digalloyl-gallagyl-hexosides, with a molecular ion at 1,085 m/z and the corresponding double charged ion [M-2H]^2-at 542 m/z; their UV-VIS spectra were coherent with such a hypothesis.

Compound 39, with molecular ion at 275 m/z, was found in the bicarbonate hydrolysates but not in the decoction and neither in AH extracts; it might be derived from the decarboxylation of EA (Hemingway and Hills, 1971).

Compounds 41, 42, and 46 were identified as isomeric forms of EA monohexoside, with a molecular fragment at 463 m/z; the presence of ion at 301 m/z was in agreement with this hypothesis, corresponding to the loss of sugar moiety. Furthermore, since compound 41 was in higher concentration than other two isobars, it was also possible to detect the dimeric [2M-H]^- ion at 927 m/z, useful for confirming its MW.

Compound 43 was hypothesized to be a derivative of granatin B, formed after the hydrolysis of one of the lactone rings, while compound 44 with a molecular ion at 951 m/z was identified as galloyl-HHDP-DHHDP-hexose or granatin B. This compound was found in decoction and bicarbonate hydrolysates, and in small amounts only in the extracts post-AH after 1 h of reaction time.

Compound 45 was identified as gallagic acid dilactone (molecular ion at 601 m/z), whose structure is reported in Figure 7. Because it was found in bicarbonate hydrolysates and not in decoction, presumably it originated after the hydrolysis of GA glycosidic esters. The compound was also found after acid extraction, resulting as the most abundant compound in solid residue along with EA.

Compound 47 was identified as EA pentoside showing a molecular ion at 433 m/z and ion at 301 m/z derived by the loss of pentoside. Compound 48 was identified as EA by comparison with pure standard. As discussed in Section 3.3, its amount in bicarbonate extracts was significantly higher than in decoction and increased with increase in NaHCO₃ concentration. In acid hydrolysates, it was found only in low amounts due to its poor solubility in acidic water, but it was the most abundant component in solid pellet produced after the hydrolysis of peel.

Compound 50, with molecular ion at 425 m/z, was identified as a decarboxylated derivative of valoneic acid dilactone (Figure 7); also as seen for compounds 19, 22, and 39, decarboxylation derivatives are only detected in bicarbonate extracts, confirming that the tannin structure tends to lose carboxyl moiety in the presence of heat and bicarbonate (Hemingway and Hills, 1971).