Potential Anticancer Activity of Pomegranate (Punica granatum L.) Fruits of Different Color: In Vitro and In Silico Evidence

Pomegranate (PMG; Punica granatum L.) fruits possess a well-balanced nutrient/phytochemical composition, with proven adjuvant benefits in experimental cancer chemotherapy; however, such bioactivity could be affected by PMG’s phenogenotype (varietal). Here, the chemical and phytochemical (UPLC-DAD-MS2) composition, antioxidant capacity and anticancer potential [in vitro (MTT assay) and in silico (foodinformatics)] of three PMG fruits of different aryl color [red (cv. Wonderful), pink (cv. Molar de Elche), and white (cv. Indian)] were evaluated. The macro/micronutrient (ascorbic acid, tocols, carotenoids), organic acid (citric/malic), and polyphenol content were changed by PMG’s varietal and total antioxidant activity (ABTS, alcoholic > hexane extract) in the order of red > pink > white. However, their in vitro cytotoxicity was the same (IC50 > 200 μg.mL−1) against normal (retinal) and cancer (breast, lung, colorectal) cell lines. Sixteen major phytochemicals were tentatively identified, four of them with a high GI absorption/bioavailability score [Ellagic (pink), vanillic (red), gallic (white) acids, D-(+)-catechin (white)] and three of them with multiple molecular targets [Ellagic (52) > vanillic (32) > gallic (23)] associated with anticancer (at initiation and promotion stages) activity. The anticancer potential of the PMG fruit is phenogenotype-specific, although it could be more effective in nutraceutical formulations (concentrates).


Introduction
Cancer is one of the five leading causes of adult death worldwide, and its burden is rising in a disproportionate manner. Data from the global cancer observatory (GLOBOCAN-GCO) indicate that~10 million deaths and~19.3 million new cases occurred in 2020, with breast (11.7%) > lung > colorectal > prostate > stomach (5.6%) cancers the most diagnosed and lung (18%) > colorectal > liver > stomach (6.9%) cancers the most lethal [1,2]. Cancer imposes the highest clinical, social, and economic burden in terms of cause-specific disability-adjusted life years (DALYs) among all human diseases [3]. While acknowledging that the most effective way of reducing cancer risk is targeting preventable risk factors (e.g., unhealthy nutrition, smoking) and correct diagnosis/control of non-preventable ones (e.g., genetic factors and primordial inflammatory response), the repositioning of OTC drugs not commonly used as chemotherapeutic aids (e.g., metformin) and the use of natural bioactive compounds (e.g., antioxidants) are effective primary/secondary aids [4,5]. It

Chemical Analysis
Triplicates analyses (CV ≤ 10%) were performed in all three dried PMG samples, using official AOAC methods for ash, protein, fat, moisture, and carbohydrate (by difference) contents.

Statistical Analysis
All continuous variables were tested for normality (Shapiro-Wilk test) and homogeneity of variance (Levene s test) and expressed as mean ± standard deviation (SD). One-way ANOVA followed by Tukey's post hoc test were carried out to evaluate any significant differences (p < 0.05) between PMG samples. Pearson product-moment correlation coefficients (r) were calculated (and depicted as heatmap and PCA plots) to detect linear relationships between antioxidant phytochemicals (phenolic and non-phenolic) and overall antioxidant capacity (hydroalcoholic and hexane extracts). Statistical differences were considered as being p < 0.05. Data analysis was performed using NCSS 2000 (NCSS Statistical Software, Kaysville, UT, USA).

and quercetin (QE) equivalents.
This composition is quite like that of USDA Food data central (FDC)-registered raw PMG (FDC ID: 169134, NDB number: 9286; https://fdc.nal.usda.gov/index.html; accessed date: 15 August 2022), and to other PMGs from India [29], and South Africa [30]. Even though the chemical composition of PMGs can be affected by several pre/post-harvest factors, even for the same phenogenotype (varietal). Its nutrient density is higher than its energy density in such a way that the fruit can be recommended for primary (healthy eating) and secondary (functional macro/micronutrient profile) prevention of many diseases, including cancer [8][9][10].

Non-Phenolic Antioxidants
One of the many mechanisms by which PMG fruit s phytochemicals exert their anticancer effects is through antioxidant activity. The total antioxidant capacity of PMG is mainly driven by phenolic species (to be discussed in the Section 3.3), ascorbic acid, carotenoids and tocols (T + T3), and other free radical scavengers with anticancer potential [5,14,23]. According to Table 1, white ( Indian ) PMG is richer in ascorbic acid, α, β-carotenes, γ-T, and β-T3 than red ( Wonderful ) and pink ( Mollar de Elche ) PMGs, but the later are better sources of lutein and total carotenoids; interesting, total tocols were found in white > pink > red PMG. Similar results in ascorbic acid content for white vs. red PMG were reported by Opara et al. [31], also agreeing with Costa et al. [32] who reported γ-tocopherol as the main vitamin E (tocol) isoform in PMG seed oils from Turkey and Israel. It is noteworthy that PMG aryls contain both seeds and juice (here combined in PMG samples) whose non-phenolic antioxidant phytochemicals contribute to the stronger hydroxide RSC and better DNA damage-preventing capacity of white PMG as compared to red PMG [33], which usually exhibits much more antioxidant capacity and phenolic content/diversity. It should be noted that, although it was not evaluated in our study, the contribution of vitamin C, carotenoids, and tocols to the total RSC should be expected to be lower than that of phenolic compounds, which are in higher concentration and diversity in the PMG samples [27].
Several epidemiological (cross-sectional/longitudinal) studies, systematic reviews and meta-analyses have supported the benefits of vitamin A (+ carotenoids), C (ascorbic acid), and E (T + T3 isoforms) in the primary prevention of certain human cancers [5]; however, evidence on the benefits of these vitamins as adjuvant chemotherapy aids remains questionable. For example, ascorbic acid improves the immunological status of patients receiving intensive chemotherapy and/or stem cell transplantations, particularly when administered parenterally rather than orally, although other effects (e.g., overall survival, clinical status, quality of life, and performance status) are not evident [34]. Carotenoid pigments are also considered potential candidates for chemoprevention and chemotherapeutics of breast, colorectal, lung, and prostate cancers and their cytotoxic/antiproliferative effect is related to a plethora of molecular mechanisms in a cancer-type manner [35]; however, such bioactivities are firstly conditioned by their first/second-pass metabolism when orally administered [5]. As for vitamin E [tocols (T + T3)], certain isoforms (γ-T, δ-T, γ-T3, δ-T3) have antiproliferative/pro-apoptotic effects on many types of cancer cells in vitro and suppress tumor progression in vivo (preclinical cancer models) by modulating key signaling pathways/cancer (e.g., eicosanoids, NF-κB, STAT3, PI3K, and sphingolipids, although such relationship is conditioned to their bioavailability) [36]. Nevertheless, although PMG aryls are not particularly rich in carotenoids and tocols, and although they are in ascorbic acid, particularly sour PMGs, their differential/synergistic contribution to the oncosuppressive activity of PMGs of different phenogenotypes remains to be evaluated [7][8][9][10].

Phenolic Compounds (PC) and Organic Acids
PC are a group of simple-to-complex molecules, characterized by having at least one phenolic ring in their structure. The bioactivity of PC in cancer is not only limited to their role in cell defense (e.g., antioxidant and anti-inflammatory activity) but also by exerting specific mechanisms at early cancer stages (initiation-promotion-transformationprogression), examples of which include cell signaling, epigenetic action, hormonal/enzyme control, immunoenhancing activity, etc. PC bioactivity depends on the chemical structure, the dose/route at which they are administered, and their synergistic/antagonistic effects with other biomolecules [5,14].
Ellagitannins and ANTO have been the most investigated PC in PMGs because of their well-known antioxidant, anti-inflammatory, cardioprotective and anticancer properties [42][43][44]. Particularly, ellagitannins [43] and ANTO [44] have potent antiproliferative and pro-apoptotic activity against several cancer cells including those of the colon, breast, lung, cervix, prostate, liver, and oral/esophageal, to name a few. Although the extractive methodology and chromatographic conditions used in this study did not allow the identification of complex ellagitannins (e.g., punicalagin), as it has been reported for "Wonderful" and "Mollar de Elche" varieties by others [37][38][39][40][41] that the presence of ellagic acid and two of its glycosides in all three varieties of PMG (pink > red, white; Table S1) accounts for this. PMG s ellagitannin-derived compounds (e.g., ellagic and gallagic acids, and urolithins A/B and B) not only have a high RSC but also antimutagenic activity [42,45]. Gallic acid and its glucosides (present in pink/white > red PMG; Table S1; see Supplementary Materials) and gallotannins are also bioactive against many cancer cells, including abnormal colonocytes [46]. Lastly, even though monomeric ANTO in white PMG fall below detection/quantification limits with the pH-differential spectrophotometric method [17], it was possible to identify them by UPLC-DAD-MS 2 , evidencing the lack of sensitivity/accuracy of the spectrophotometric method.

Antioxidant Capacity
As previously stated, antioxidant phytochemicals play an outstanding role in chemoprevention [5], including DNA-protective capacity [33], as well as in enhancing antiinflammatory and other cytoprotective effects [8,11,42,44]. According to Figure 1, the radical assay sensitivity to detect differences between samples was ABTS > DPPH > ORAC in methanolic (hydrophilic antioxidants) hexane (hydrophobic antioxidants) extracts; the ABTS − radical was efficiently scavenged (µmol TE/g dw) by red (44. with no juice), but not in juice. Similar results were also reported for Spanish cultivars, which also include "Wonderful" and "Mollar de Elch"' [40]. Pearson's product-moment correlation analysis ( Figure S1 in the Supplementary Materials), evidenced a positive strong-to-moderate trend (r ≥ 0.74) between ABTS-M and total PC, flavonoids, carotenoids contents, and DPPH-M. However, ABTS-M demonstrated an inverse relationship (r ≥ −0.73) with ascorbic acid and tocol content, as well as the scavenging ability of DPPH and ABTS▪− radicals for hexanoic extract (DPPH-H and ABTS-H). Conversely, ABTS-H a showed strong positive (r ≥ −0.88; ascorbic acid, tocol and DPPH-H) and moderate (r ≥ −0.61) relationship with the rest of the assayed parameters. Yang et al. [50] studied thirteen PMG varieties cultivated in China (Zaozhuang), reporting a positive correlation between antioxidant capacity (DPPH radical) and total contents of PC, flavonoids, and hydrolyzable tannins (r ≥ 0.89) in both the pericarp (peel) and seeds (aryls with no juice), but not in juice. Similar results were also reported for Spanish cultivars, which also include "Wonderful" and "Mollar de Elch"' [40].
The aforementioned apparent causal relationships (↑ [antioxidant species]: ↑ RSC; ↑ [hydrophilic antioxidants]: ↓ [lipophilic antioxidants]) are partially explained by the individual antioxidant capacity (radical-specific quenching) and the relative abundance of each antioxidant phytochemical extracted under hydrophilic (methanol-water; LogPow ~1.0) and hydrophobic (hexane; LogPow = 3.9) conditions [27]. Unlike DPPH and ABTS▪− radicals-based assays, the ORAC assay is considered more convenient to evaluate chainbreaking antioxidant activity within biological systems where reactive oxygen (ROS) and nitrogen species (NOS) are more relevant [20]. ROS plays a dual role in cancer by promoting cell proliferation-survival-adaptation to hypoxia and by triggering oxidative-stress cancer cell death, both cases handled with ROS-manipulation therapies [51]. However, this assay was less sensitive to discriminate PMG phenogenotypes, a fact that deserves a more detailed study aimed to uncover the fine-tuning mechanism exerted by each antioxidant specie.

In Silico Gastrointestinal Fate Prediction
Machine learning/artificial intelligence (ML/AI) algorithms have recently gained attention within the pharma/nutraceutical industry and are being applied at different stages of the natural product drug research pipeline [52], including those derived from foods. This staging research process is initially fed with either chemical structure data coming from high-through output chemical characterization methods (e.g., GC-MS, HPLC-MS) or by computer-assisted data mining of natural products [53]. The next step is to process and filter large chemical databases by using cheminformatic platforms, aiming to explore the bioactive potential of each molecule (targeted/untargeted approaches) and artificially ranking its target bioactivity and applications. Particularly, ADME/Tox (physicochemical  [27]. Unlike DPPH and ABTS − radicals-based assays, the ORAC assay is considered more convenient to evaluate chainbreaking antioxidant activity within biological systems where reactive oxygen (ROS) and nitrogen species (NOS) are more relevant [20]. ROS plays a dual role in cancer by promoting cell proliferation-survival-adaptation to hypoxia and by triggering oxidative-stress cancer cell death, both cases handled with ROS-manipulation therapies [51]. However, this assay was less sensitive to discriminate PMG phenogenotypes, a fact that deserves a more detailed study aimed to uncover the fine-tuning mechanism exerted by each antioxidant specie.

In Silico Gastrointestinal Fate Prediction
Machine learning/artificial intelligence (ML/AI) algorithms have recently gained attention within the pharma/nutraceutical industry and are being applied at different stages of the natural product drug research pipeline [52], including those derived from foods. This staging research process is initially fed with either chemical structure data coming from high-through output chemical characterization methods (e.g., GC-MS, HPLC-MS) or by computer-assisted data mining of natural products [53]. The next step is to process and filter large chemical databases by using cheminformatic platforms, aiming to explore the bioactive potential of each molecule (targeted/untargeted approaches) and artificially ranking its target bioactivity and applications. Particularly, ADME/Tox (physicochemical properties) screening + untargeted bioactivity prediction (pharmaceutical interest) is probably the most useful cheminformatic approach in the nutraceutical field [23,24].
Bioavailability radars offer preliminary insights into the drug-likeness of natural bi oactive compounds [23], and the structural feature distinguishing GI-permeants from non-permeants is the low fraction of carbons in the sp 3 hybridization state. However, non of these sixteen molecules can cross the BBB (Figure 3). According to Daina et al. [23] smal molecules (MW, 150-500 g/mol) with low TPSA (20-130 Å 2 ), solubility (LogS ≤ 6.0), and flexibility (≤9 rotatable bonds), yet a high fraction (≥0.25) of sp 3 carbons, are more likely to permeate GI/BBB. In this study, phenolic acids (aglycones) but not their glycosylated deri vates (bioavailability score= 0.17-0.55, synthetic accessibility 3.79-5.32) complied with these chemical features. Phenolic acids (PC bearing at least one phenol moiety within a resonance stabilized structure) are efficiently extracted from food matrices within the G tract (bioaccessibility) and effortlessly absorbed into the intestine, reaching systemic cir culation quite rapidly [53]. Under simulated (in vitro) gastrointestinal conditions (static in vitro digestion), PMG phenolic acids are more stable in gastric and intestinal condition than ANTO and Punicalagin A (ellagitannin) [54]. Moreover, glycosylated derivates and complex forms of phenolic acids (ellagitannins and gallotannins) can be efficiently hydro lyzed by pancreatic (hydrolases) or microbial (esterases) enzymes into their monomeri forms, which are further biotransformed into cinnamic (ellagic acids)/ benzoic (phenoli acids) metabolites of much lower MW yet higher bioactivity, increasing their odds of BBB permeation [55]. The bioaccessibility and antioxidant activity of catechins (including D (+)-catechin) positively correlates with their binding affinity to dietary (e.g., β-lactoglobu lin, and β-casein) and GI proteins (e.g., P-gp) under simulated GI conditions, a fact partic ularly important in colon cancer chemotherapy [56,57].

Protein-Targetting Prediction
Almost every natural product has special chemical properties which endow it with the capacity to ameliorate various diseases. Particularly, PMG phytochemicals exert antioxidant, anti-inflammatory, and hypoglycemic effects associated with a plethora of non-communicable diseases, although their oncosuppressive action seems to be cancerspecific [58]. On the latter poiny, the antiproliferative (cell cycle arrest at S-G2-M), proapoptotic, free-radical inhibition/trafficking, antiangiogenic, and antimetastatic (antimigration/invasion) qualities of PMG's PC have been studied in detail for breast, lung, thyroid, colon, and prostate cancer [7][8][9][10]. In this study, potential protein targets in cancer (initiationpromotion-transformation-progression) and other metabolic intermediaries were predicted with SwissTargetPrediction 2019. This is an open web tool that calculates the similarity threshold of a given small molecule with other molecules compiled in curated and cleansed collections of known bioactive compounds (reverse screening) [59], and those with presumably anticancer potential were confirmed with UniProt [25] and Pharos [26] databases. The number of cancer-related protein targets was PMG-specific [red (129) > pink (109) > white (55)], as related to their phytochemical fingerprints (Figure 4)  predicted with SwissTargetPrediction 2019. This is an open web tool that calculates the similarity threshold of a given small molecule with other molecules compiled in curated and cleansed collections of known bioactive compounds (reverse screening) [59], and those with presumably anticancer potential were confirmed with UniProt [25] and Pharos [26] databases. The number of cancer-related protein targets was PMG-specific [red (129) > pink (109) > white (55)], as related to their phytochemical fingerprints (Figure 4)   and other (Y-axis) protein target prediction for all identified phytochemicals (see Figure 2 for coded phytochemicals) in white (circle), pink (cubes) and red (pyramids) PMGs. Protein targets were identified (SwissTargetPrediction 2019 [59]; http://www.swisstargetprediction.ch/; accessed date: August 15 th , 2022) and confirmed with UniProt [25] and Pharos [26]).
It is noteworthy that the number of protein targets against cancer followed the same trend as for other targets involved in other cellular processes, both in bioavailable (all but D-(+)-catechin) and non-bioavailable phytochemicals (Figure 4). Bioavailable phenolic acids [ Figure 3; ellagic (1), gallic (4), and vanillic (7) acids)] have demonstrated several anticancer mechanisms in vitro, ex vivo, and in vivo [53,60,61].
Lastly, among other protein targets relevant in the clinical course of cancer, ellagic acid, gallic acid, and vanillic acid to a lesser extent, have a strong inhibitory activity on several human carbonic anhydrase (hCA, metalloenzyme; E.C. EC 4.2.1.1) isoforms (I, II, II, IV, VB, VI, VII, IX, XII, XIV). hCAs are strategically distributed in human organs and cells (cytosolic, membrane-associated, fluids) and are associated with key physiological activities; particularly, within the microenvironment of tumor cells, membrane-bound hCAs IX and XII activate the hypoxia-inducible factor 1 and 2 (HIF-1/2) consequently activating the expression of several genes involved in glucose metabolism, angiogenesis, and pH regulation [62][63]. Since natural products bearing phenolic rings are potent inhibitors of hCAs (in an isoform-specific manner) [64] and both ellagic and gallic acids have a catechol A-PAINs alert (Table S2; see Supplementary Material), this explains their promising potential as hCA inhibitors.
It is noteworthy that the number of protein targets against cancer followed the same trend as for other targets involved in other cellular processes, both in bioavailable (all but D-(+)-catechin) and non-bioavailable phytochemicals (Figure 4). Bioavailable phenolic acids [ Figure 3; ellagic (1), gallic (4), and vanillic (7) acids)] have demonstrated several anticancer mechanisms in vitro, ex vivo, and in vivo [53,60,61].
Lastly, among other protein targets relevant in the clinical course of cancer, ellagic acid, gallic acid, and vanillic acid to a lesser extent, have a strong inhibitory activity on several human carbonic anhydrase (hCA, metalloenzyme; E.C. EC 4.2.1.1) isoforms (I, II, II, IV, VB, VI, VII, IX, XII, XIV). hCAs are strategically distributed in human organs and cells (cytosolic, membrane-associated, fluids) and are associated with key physiological activities; particularly, within the microenvironment of tumor cells, membrane-bound hCAs IX and XII activate the hypoxia-inducible factor 1 and 2 (HIF-1/2) consequently activating the expression of several genes involved in glucose metabolism, angiogenesis, and pH regulation [62,63]. Since natural products bearing phenolic rings are potent inhibitors of hCAs (in an isoform-specific manner) [64] and both ellagic and gallic acids have a catechol A-PAINs alert (Table S2; see Supplementary Material), this explains their promising potential as hCA inhibitors.

In Vitro Citototoxicity Assays
The cytotoxic activity (MTT assay) of PMG samples against normal (retinal) and cancer (breast, lung, colorectal) cell lines is depicted in Figure 5.

Conclusions
In this study, the in vitro assays indicated that the antioxidant capacity (ABTS, alcoholic > hexane extract) and phytochemical fingerprint [ascorbic acid, tocols, carotenoids, organic acids, and PC (relative abundance/chemical nature)] of PMGs is phenogenotype-specific [red > pink > white], although their in vitro cytotoxicity was the same (IC 50 > 200 µg.mL −1 ) against normal (retinal) and cancer (breast, lung, colorectal) cell lines. On the other hand, chemometric analyses showed that ellagic (pink) > gallic (white) > vanillic (red) acids may interact with multiple molecular targets in the cancer continuum (initiation-progression). So, by employing a comprehensive and multidisciplinary stepwise experimental design [in vitro assays (several cell lines, high-throughput chemical characterization) + in silico studies (foodinformatics)], this study documents the differential anticancer potential of three PMG fruits of different flesh (aryl) colors [red (cv. Wonderful), pink (cv. Molar de Elche), white (cv. Indian)], although such anticancer potential could be more effective in nutraceutical formulations such as fruit concentrates. The practical application of the evidence reported here may result in the formulation of novel PMG fruit-based OTC nutraceuticals, highlighting their phenogenotype (varietal)-specific phytochemical value as a market differentiation element, increasing the PMG fruit market share currently held by PMG ('Wonderful') dry concentrates producing reconstituted juices and medicinal syrups.