Biology Assessing the anticancer e ﬀ ects associated with food products and/or nutraceuticals using in vitro and in vivo preclinical development-related pharmacological tests

This review is part of a special issue entitled “ Role of dietary pattern, foods, nutrients and nutraceuticals in supporting cancer prevention and treatment ” and describes a pharmacological strategy to determine the potential contribution of food-related components as anticancer agents against established cancer. Therefore, this review does not relate to chemoprevention, which is analysed in several other reviews in the current special issue, but rather focuses on the following: i) the biological events that currently represent barriers against the treatment of certain types of cancers, primarily metastatic cancers; ii) the in vitro and in vivo pharmacological pre-clinical tests that can be used to analyse the potential anticancer e ﬀ ects of food-related components; and iii) several examples of food-related components with anticancer e ﬀ ects. This review does not represent a catalogue-based listing of food-related components with more or less anticancer activity. By contrast, this review proposes an original pharmacological strategy that researchers can use to analyse the potential anticancer activity of any food-related component — e.g., by considering the crucial characteristics of cancer biological aggressiveness. This review also highlights that cancer patients undergoing chemotherapy should restrict the use of “ food complements ” without supervision by a medical nutritionist. By contrast, an equilibrated diet that includes the food-related components listed herein would be bene ﬁ cial for cancer patients who are not undergoing chemotherapy.


Introduction
Cancer imposes an enormous burden on societies in more and less economically developed countries alike. The occurrence of cancer is increasing due to the growth and ageing of the population, as well as the increasing prevalence of established risk factors, such as smoking, being overweight (relating to abnormal and/or inappropriate food consumption), physical inactivity, and changing reproductive patterns associated with urbanization and economic development [1]. Torre et al. [1] recently reviewed the cancer incidence worldwide, while Jemal et al. [2] performed this analysis for US cancer patients [2].
For Hanahan and Weinberg [3], the multistep development of human tumours includes sustained proliferative signalling, evasion from growth suppressors, resistance to cell death, the capacity of replicative immortality, induction of angiogenesis, activation of invasion and metastasis, the reprogramming of energy metabolism and evasion from immune destruction. These authors further report that cancers exhibit another dimension of complexity because they contain a repertoire of recruited, ostensibly normal cells that contribute to the "tumour microenvironment" [3]. Thus, cancerous tissue is highly heterogeneous, and this heterogeneity contributes to the ineffectiveness of current chemotherapy agents. The current review highlights that several food-related components can hinder the biological development of cancer as described by Hanahan and Weinberg [3].

Generalities
Modern chemotherapy is mainly based on the use of cytotoxic or targeted agents [4], usually applied after surgery and after or along with radiotherapy, to which immunotherapy can also be added. Most cytotoxic anticancer agents are of natural origin, while targeted therapies generally result from computer modelling followed by synthesis. However, both types of therapy, unfortunately, display major limitations with respect to cancer cell heterogeneity [4]. The major limiting factor for the use of cytotoxic drugs is their toxicity towards many healthy organs. However, targeted agents exhibit the same frequency and severity of toxicities as traditional cytotoxic agents, with the main difference being the nature of the toxic effects-e.g., alopecia, myelosuppression, mucositis, nausea, and vomiting for cytotoxic therapies versus vascular, dermatologic, endocrine, coagulation, immunologic, ocular, and pulmonary toxicities for targeted therapies [4]. Immunotherapy is also associated with limiting toxic effects [5,6]. Barber and colleagues [7] accordingly reported that understanding subclonal heterogeneity architectures and cancer evolution processes is critical for the development of effective therapeutic approaches, which can control or thwart cancer evolutionary plasticity. Arnedos et al. [8] stated that, although several successfully targeted agents have been developed in recent years, most tumours eventually develop drug resistance, potentially due to intratumoural heterogeneity and the selection of additional biochemical events. Ramos and Bentires-Alj [9] explained that the plasticity of tumour cells leads to the development of drug resistance by distinct mechanisms, including the following: (i) mutations in the target, (ii) reactivation of the targeted pathway, (iii) hyperactivation of alternative pathways, and (iv) cross-talk with the microenvironment. Dorel et al. [10] stated that signalling pathways implicated in cancer create a complex network with numerous regulatory loops and redundant pathways and that this complexity also explains the frequent failure of the one-drug-one-target paradigm of treatment, resulting in drug resistance in patients. These authors proposed that cancer treatment should be extended to a combination of therapeutic approaches to overcome the robustness of the cell signalling network [10]. As highlighted in the following section, the consumption of certain types of food components/diet regimens can help as an added weapon to combat certain types of cancer, in addition to conventional treatments. Indeed, many food-related components display distinct mechanisms of action in terms of anticancer effects. Thus, a diversified and equilibrated diet can be deleterious to several subpopulations within a given cancer as detailed in the following sections. However, it must also be emphasized that certain diet components can be deleterious for cancer patients with certain ongoing types of chemotherapy (as summarized at the end of the current review).
In addition to the mechanisms described above, cancer cellular drug resistance can also be associated with the altered expression of the ATPbinding cassette (ABC) family of transporters, the most common cause of multidrug resistance (MDR), alterations of DNA repair pathways, and resistance to pro-apoptotic stimuli [11]. Finally, the ineffectiveness of cancer chemotherapy is associated with the tumour microenvironment, hypoxia and the development of metastases, as detailed in the following sections. An anticancer agent that impairs the biology of the tumour microenvironment represents a promising and very innovative anticancer drug. This is the case, for example, for the tunicate metabolites trabectedin (marketed as Yondelis) and plitidepsin (in phase III clinical trials) [12].

Resistance to pro-apoptotic stimuli
The link between the evasion of apoptosis and cancer development is implicitly clear if one considers how many cells are produced each day and, hence, how many cells must die to make room for the new ones. Cells frequently experience noxious stimuli that can cause lesions in their DNA. These lesions need to be repaired efficiently, or, in the case of irreparable damage, the cell must be killed to prevent the subsequent division of aberrant cells that may fuel tumourigenesis. As reported by Kelly and Strasser [13], the detection of genetic lesions in human cancers that activate pro-survival genes or disable pro-apoptotic genes serves as direct evidence that defects in apoptosis can cause cancer. Evasion of apoptosis is thus a requirement for both neoplastic transformation and the sustained growth of cancer cells [13,14]. Mohammad et al. [14] remind us that most anticancer therapies trigger apoptosis and related cell death networks to eliminate malignant cells. However, deregulated apoptotic signalling, particularly the activation of anti-apoptotic systems, allows cancer cells to escape this programme, leading to tumour survival, therapeutic resistance and cancer recurrence. In other words, a promising anticancer drug should be a compound that kills cancer cells through the activation of non-apoptosisrelated cell death pathways [15]. Mohammad et al. [14] recently reviewed the key apoptosis-resistance targets that include the following: (i) B-cell lymphoma 2 (Bcl-2) and myeloid cell leukaemia 1 (Mcl-1) proteins, (ii) autophagy processes; (iii) necrosis and necroptosis, (iv) heat shock protein signalling, (v) the proteasome pathway; (vi) epigenetic mechanisms, and (vii) aberrant nuclear export signalling.

Cancer stem cells
Similar to normal tissue, many tumours have a hierarchical organization where tumourigenic cancer stem cells (CSCs) differentiate into non-tumourigenic progenies [26,27]. Stem cells are often localized to hypoxic niches within tissues, and hypoxia-inducible factors (HIFs) play key roles in the maintenance of pluripotent and multipotent stem cells, as well as CSCs, which are also known as tumour-initiating cells [27]. Islam et al. [28] and Adorno-Cruz et al. [29] reviewed the roles of CSCs in the metastatic process, treatment resistance, and cancer recurrence via the activation of different signalling pathways, such as Notch, Wnt/ β-catenin, transforming growth factor-β (TGF-β), Hedgehog, PI3 K/ Akt/mTOR and JAK/STAT. Cojoc et al. [26] accordingly reported that strategies based on the combination of conventional therapies targeting bulk tumour cells and CSC-specific pathways bear significant promise to improve cancer treatment outcomes compared with monotherapies. Marucci et al. [30] recently reviewed the ability of 49 different natural products to influence CSC biology.

Hypoxia
The tumour microenvironment exerts a complex and strong influence on the tumour cell phenotype [31]. Hypoxia represents one of these tumour microenvironmental effects on cancer cell biology, and it occurs, for example, with poor tumour neoangiogenesis and increased oxygen consumption [31,32]. As also emphasized by Span and Bussink [31], hypoxia is a multifactorial phenomenon involving oxygen tensions ranging from < 0.01% (anoxia) to 5% and can be chronic, acute, or cycling, all with differential effects on tumour cells. When cancer cells face hypoxic conditions, they activate intracellular signalling pathways, including HIF-1-mediated gene expression, the unfolded protein response, and AKT-mammalian target of rapamycin signalling [31]. Activation of these pathways, in turn, induces aggressive, metastatic and treatment-insensitive tumours [31][32][33][34][35]. Unwith et al. [36] accordingly reported that HIF-1 overexpression in many aggressive cancer types, as well as its role in the establishment of metastatic disease and treatment resistance, makes it an attractive potential target for cancer drug development. Irregular blood flow and large distances between functional blood vessels lead to poor drug distribution within solid tumours [37]. Cells that are distal from functional blood vessels are not exposed to effective concentrations of the drug, resulting in therapeutic resistance. metastatic disease [23][24][25]. The metastatic cascade is a very complex process [23,[38][39][40]] that schematically begins with the movement of cancer cells from a primary site to other locations, and this process, therefore, requires at least the ability of metastatic cancer cells to overcome anoïkis (to resist pro-apoptotic stimuli), invade neighbouring tissues, intravasate and then extravasate blood or lymphatic vessels, and evade detection by the immune system. The additional complexity of the metastatic cascade results from metastases arising from residual disseminated tumour cells years after primary tumour treatment because these residual cells can enter dormancy and evade therapies [39,41]. Tumour cell dormancy is regulated by normal stem cell quiescence, extracellular and stromal microenvironments, autophagy and epigenetics [39,41]. Next, the epithelial-mesenchymal transition (EMT), which is a hallmark of normal embryonic processes; becomes dysregulated in some cancer cells; and leads to dysfunctional cell-cell adhesive interactions, loss of cell-cell junctions, and restructuring of the cytoskeleton [42]. These events collectively result in the loss of apical polarity and acquisition of a more spindle-shaped morphology, dramatically facilitating cancer cell migration [42,43]. EMT is also associated with resistance to cancer therapies and CSC maintenance [42].
The metastatic process also leads to marked modifications in the cytoskeleton organization at the level of both actin (microfilaments) and tubulin (microtubules). Several anti-tubulin compounds of natural origin have already been marketed (e.g., paclitaxel, docetaxel, and vincristine) and/or have reached phase II and phase III clinical trials [44-46].

Traits of compounds with anticancer effects in the preclinical stages of development
A potential anticancer compound, such as a food-related component, before being labelled "promising", should provide positive hits against several of the biological events we detailed above. By contrast, researchers claiming that a given compound is a "promising" anticancer agent with data based only on in vitro 3-(4,5-dimethylthazol-2-yl)-2,5diphenyltetrazolium bromide (MTT) colourimetric-based measurements are actually overselling their data. We provide our view regarding this point below.

In vitro
A large majority of research articles' claims regarding "promising" anticancer effects for various types of compounds, especially for natural  [14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32] compounds, including many food-related components, present i) only "cytotoxic" effects ii) that were measured using a colourimetric assay (such as the popular MTT assay) and iii) using very few cancer cell lines (in general 1-3), with most of them displaying high sensitivity to proapoptotic stimuli. However, we disagree with claims based on such techniques. First, cytotoxicity is not a synonym for growth inhibitory effects. Cytotoxicity means "cell-killing effects", while "growth-inhibitory effect" is self-explanatory, and such an effect can result from cell cycle blockade, the induction of cell death (e.g., cytotoxicity, which relates to apoptosis or other cell death types) and cell detachment from the multiwell bottom (thus, anti-adhesive, i.e., anti-metastatic effects). Second, a colourimetric assay, such as the MTT assay, does not measure direct cell killing effects, i.e., cytotoxicity, if one considers the GI 50 (or popularly referred to as the IC 50 ) concentration only. The MTT colourimetric assay indirectly assesses, using spectrophotometric analyses, the number of metabolically active cells that can transform the yellow substrate MTT into the blue formazan dye via mitochondrial reduction involving succinate dehydrogenase. The GI 50 index obtained using the MTT colourimetric assay for a compound of interest solely indicates that at a given concentration and for a given period of cell culture, this compound reduces the growth of the particular cell line by 50%. The term "cytotoxic" can be employed for a given compound if, and only if, experimental evidence of cell death is demonstrated, for example, using cellular vital dyes (e.g., Trypan blue or propidium iodide) or enzyme release (such as lactate dehydrogenase, LDH) in the culture medium [47]. Morphological analyses can also be helpful in identifying cytotoxic [48] versus cytostatic [49] anticancer compounds. The in vitro strategy set up by the US National Cancer Institute (NCI) on the 60-cancer-cell-line panel can also make a clear difference between cytotoxic and cytostatic growth inhibitory effects, while using colourimetric assays (but not using only the GI 50 concentration) as comprehensively explained by Shoemaker [50]. However, false-positive and false-negative data can also be obtained with colourimetric assays for various types of compounds [51-53], especially in the case of polyphenols [54], which represent the largest group of food-related components with anticancer effects (as highlighted in Tables 1 and 2).
We report in Table 2 12 "descriptors", i.e., 12 cancer-related characteristics of biological aggressiveness. "Bioselectivity" (descriptor 1 in Table 2) is not a synonym for "selectivity". Bioselectivity is a term that refers to the differential cytotoxic/cytostatic/growth inhibitory effects observed for a given compound at a given concentration between normal and cancer cells. An ideal bioselectivity index in vitro would be > 10 for the mean GI 50 concentrations obtained between at least 3 normal and 3 cancer cell lines [55,56]. The in vitro bioselectivity index provides a more than rough estimation regarding the potential in vivo therapeutic index. The term "selective" can be used when a given compound displays in vitro growth inhibitory selectivity with respect to a given histological group(s) of cancer cell lines out of a larger number of histological cancer types tested for the same compound. This notion of selectivity can be studied with the help of the NCI's 60-cell-line panel [48,50]. Using cancer cell lines that belong to this panel also enables any type of compound (including food-related components) to be analysed in terms of genomic [ In terms of cancer cell death activation, a "promising" anticancer compound should kill apoptosis-resistant cancer cells (descriptor 2; Table 2) without activating any component of MDR signalling pathways [48,56] (descriptor 3; Table 2). We label descriptor 4 in Table 2 for those compounds that can kill cancer cells via non-apoptotic pathways or that induce apoptosis quite indirectly. Non-apoptotic-related cell death pathways include paraptosis, irreversible autophagy, irreversible senescence, necroptosis, ferroptosis, methuosis, parthanatos, mitotic catastrophe, oncosis, pyroptosis, and autosis [15]. Indirect induction of apoptosis in cancer cells includes, for example, the targeting of ion channels/transporters [48] and specific sets of kinases as already demonstrated for certain polyphenols [63,64].
A "promising" anticancer compound should also kill CSCs (descriptor 5; Table 2). In addition, a "promising" anticancer compound should selectively kill tumour endothelia under hypoxic conditions only (descriptor 6; Table 2) because the tumour microvasculature in tumour hypoxic regions physiologically differs from normal tissue microvasculature [65][66][67]. In addition, the very popular human umbilical vein endothelial cell (HUVEC) model relates to endothelial cells from normal large vessels and not at all to endothelial cells from tumour microvessels. Many articles from the scientific literature make conclusions regarding "promising antiangiogenic effects" for food-related components assayed on HUVECs cultured under normoxic conditions. Descriptor 7 in Table 2 relates to scientific references that report the use of tumour microvessel-related cells when assaying a food-related component.
A "promising" anticancer compound should target several components of the tumour microenvironment, and the use of in vitro 3D models could help in identifying promising anticancer compounds that impair the physiology and biology of the tumour microenvironment [68][69][70][71] (descriptor 8; Table 2). This is absolutely not the case with conventional 2D cell cultures. In addition, the use of freshly obtained cancer cells from cancer patients [72,73] (descriptor 9; Table 2) as well as circulating tumour cells [74,75] (descriptor 10; Table 2) should also be used to validate those data obtained on permanent cancer cell lines, some of which were established in vitro decades ago.
Finally, what really constitutes a "promising" antimetastatic compound from an in vitro point of view? While it is the metastatic process of cancer cells escaping the primary tumour site to colonize organs at a distance that kills cancer patients, it appears unlikely to us that a given compound claimed as displaying "marked in vitro antimetastatic effects" in many publications would actually block the escape of cancer cells from closed plastic flasks (e.g., a primary organ in which a cancer develops) to invade other closed flasks (e.g., the organs invaded during the metastatic process) in the incubator. Therefore, we have serious doubts regarding overly optimistic claims relating to antimetastatic effects obtained in vitro and that only include i) the partial destruction of cytoskeleton components (e.g., tubulin), ii) a reduced level of migration of a pure cancer cell population prisoner of a closed plastic flask, or iii) pseudoinvasion features relating to the measurements of protease inactivation and/or transwell invasion membranes. The antimetastatic properties of a given anticancer compound must be determined in appropriate in vivo models. However, several models and techniques are already available to really assess the in vitro potential antimetastatic characteristics of various types of compounds. These models and techniques refer to phenotypic analyses [76][77][78][79][80]. Nevertheless, we report in the 11th column of Table 2 those references that describe sound "antimetastatic" data obtained in vitro.

In vivo
Many authors claim that they have identified a "promising" anticancer compound because they observed significant tumour growth decrease in vivo. If such in vivo data are obtained with subcutaneous (s.c.) models [whether using a syngeneic (mouse cancer grafted in mice) model or a xenograft (human cancers grafted in mice) model], we suggest that the compound in question is not promising because, in the clinical setting, s.c. cancers [whether primary or secondary (metastatic)] are cured by surgery. To be labelled as promising from an in vivo point of view, a compound must be tested in orthotopic models (whether syngeneic or xenografted), leading to the development of metastases [81] or, in the worst case, to pseudometastases [82]. Descriptor 12 in Table 2 relates to in vivo testing with orthotopic models. In addition, we ignored any data obtained in vivo with the mouse P388 leukaemia model. Indeed, survival increases induced by a compound of P388 Table 2 A rational tool to evaluate the potential anticancer effects of food-related components in preclinical in vitro and in vivo testing.  - [188,189] -Cinnamic acid [190] -- [191,192] [193] -- [192] -- [194] [192] Cordycepin - [195] [195] [196,197] [198] ----- [96,196] [196,197,199] Curcumin [200,201] [ The compounds are listed in alphabetical order. b We have selected 12 pharmacological characteristics ("descriptors") to determine the potential of a food-related component as an anticancer agent. The numbering is as follows: 1 = the compound is bioselective; 2 = the compound kills apoptotic-resistant cancer cells directly (as a single agent) or indirectly (in combination with another treatment); 3 = the compound overcomes the MDR phenotype; 4 = the compound kills cancer cells via non-apoptotic cell death pathways (in addition to pro-apoptotic ones); 5 = the compound impairs the growth and/or migration of CSCs; 6 = the compound could be antiangiogenic under hypoxic conditions; 7 = the antiangiogenic effects were determined on tumour microvesselrelated or on progenitor endothelial cells; 8 = the compound impairs the tumour microenvironment; 9 = the compound impairs the growth of human cancer primoculture (not only established cell lines); 10 = the compound is active against circulating tumour cells; 11 = the compound displays antimetastatic effects in vitro; 12 = the compound is active in vivo against metastatic features in orthotopic cancer models.
leukaemia-bearing mice are of little value because the P388 leukaemia model is, by far, too chemosensitive [83], and it was abandoned by the NCI in the mid-1980s because of its exacerbated chemosensitivity [50].

Contribution of nature in the fight against human cancers
According to the literature analyses of Newman and Cragg [84,85], over the timeframe from the 1940s to the present, approximately 49% of all approved drugs are either natural products or their direct derivatives. Furthermore, their role in the drug discovery process is especially pronounced in the anticancer and infectious disease areas, where the fractions of the drugs derived from natural products amount to 60% and 75%, respectively [85].
The majority of food-related components that are beneficial to human health are of natural origin; some of them display anticancer effects as highlighted in the following sections. However, the diet profiles of people from Africa, North America, South America, Asia, Australia, and Eastern, Western, North and South Europe markedly differ. There are thousands food-related components of natural origin that can more or less impair cancer progression, and these components can originate from plants, macroscopic mushrooms, insects, and terrestrial (other than insects) and marine invertebrates. From more than a gross picture, diets based on a high consumption of "plants" (vegetables and fruits), fish and insects and a low consumption of alcohol and mammalian fat apparently lead to a lower cancer incidence than diets based on a weak consumption of plants and a high consumption of red meal, mammalian fat and alcohol. Table 1 describes some examples of food-related components with anticancer activity. Importantly, approximately 57,000 publications are available in the PubMed database (as of December 2016) with respect to the keywords "food AND cancer". Therefore; we present to the readers of the current review a "grid" ( Table 2) that is easy to read and understand; enabling each reader; especially young researchers with limited experience in the cancer field; to draw their own conclusions on a given food-related component (perhaps not listed in Table 1). Additional descriptors could include; for example; cancer cell epigenetics and cell cycle blockade (that will finally lead to apoptosis) [14]. If a given food-related component is associated with such an anticancer effect; we reported this effect in the 4th column of Table 2. We have also limited ourselves to a maximum of 3-4 references (the most recent ones) for a given compound and a given cancer descriptor in Table 2. For some descriptors and a given food-related component; more than dozens of references exist. Table 1 contains examples of food-related components selected to illustrate their anticancer effects according to the descriptors used in Table 2.

Examples of food-related compounds with anticancer activity
There are ∼8000 chemical compounds found in the human diet today that are identified as dietary polyphenols [86]. The most common polyphenols include phenolic acids, anthocyanidins, flavonoids and tannins. Table 2 shows that several polyphenols impair crucial steps of cancer biology.
As highlighted in the Introduction, three crucial characteristics of innovative anticancer agents relate to their capability in i) killing cancer cells resistant to pro-apoptotic stimuli (such as metastatic cancer cells), ii) killing cancer cells through non-apoptotic-mediated cell death pathway, and iii) actually impairing the biological behaviour of CSCs. We mention below in the text whether each compound under discussion possesses or not one, two or three of these characteristics.
Curcumin is the most studied polyphenol, and the PubMed database contains ∼9500 publications relating to this compound with ∼1100 reviews. This compound is a very good example of how difficult an exhaustive synthesis would be of all of the data already published in the literature for a given compound such as curcumin and the reason why we have chosen to propose a "grid" ( Table 2) to summarize part, and only part, of these data. The synthetic analysis proposed here for curcumin clearly highlights that this compound displays marked anticancer effects, and it is well known that the consumption of a curcuminrich diet is inversely correlated with several human cancer types [97,206]. Its effectiveness as an anticancer drug is hampered by its poor oral bioavailability, but many approaches are currently being developed to solve this problem. We do not discuss this point in detail here because it is beyond the focus of this review. In addition to the references we cite in Table 2, it is well known that curcumin inhibits many cell signalling pathways in cancer cells such as the nuclear factor-kappa B (NF-κB), activator protein 1 (AP-1), cyclooxygenase-2 (COX-2), matrix metalloproteinases (MMPs), cyclin D1, epidermal growth factor receptor (EGFR), Akt, β-catenin, and tumour necrosis factor (TNF) [388][389][390]. Curcumin kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptotic-related cell death pathways in cancer cells, impairs the biology of CSCs, and is bioselective ( Table 2).
Anthocyanidins, the sugar-free counterparts of anthocyanins, include aurantinidin, capensinidin, cyanidin, delphinidin, europinidin, hirsutidin, malvidin, pelargonidin, peonidin, petunidin, pulchellidin and rosinidin [86]. These compounds are common plant pigments that are mainly found in berries, which also contain many polyphenols [86]. The protective roles of dietary berries in cancer have recently been reviewed by Kristo et al. [86]. Table 2 shows that anthocyanidins impair most of the biological processes involved in cancer aggressiveness including cancer chemoresistance.
Green tea is one of the most extensively consumed beverages in the world (along with coffee). Most of the anticancer effects of green tea are related to epigallocatechin 3-gallate (EGCG) [391,392]. There are hundreds of publications related to the anticancer effects of EGCG. The few ones we cite in Table 2 clearly show that EGCG impairs most of the biological characteristics associated with cancer aggressiveness. Additionally, among other effects, EGCG kills cancer cells resistant to proapoptotic stimuli, induces non-apoptotic-related cell death pathways in cancer cells, impairs the biology of CSCs, and is bioselective (Table 2).
Quercetin, the most common flavonol in berries (Table 1), exerts anti-inflammatory effects by acting on inflammatory mediators such as IL-6, IL-8, IFN-γ, iNOS, COX-2 and TNF-α [393,394]. Many studies have described the pro-apoptotic effects induced by quercetin in various cancer cell lines [393,395,396]. However, the cancer cell lines used in these studies are sensitive to pro-apoptotic stimuli, in contrast to chemoresistant cancer, as detailed in the Introduction. Table 2 shows that quercetin exerts anticancer effects through the targeting of many biological events directly implicated in cancer aggressiveness. Fisetin is structurally similar to quercetin and is present in many edible fruits and vegetables (Table 1). Fisetin also displays potent anticancer activity ( Table 2). Quercetin kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptotic-related cell death pathways in cancer cells, impairs the biology of CSCs, and is bioselective ( Table 2).
Apigenin is a natural flavone present in many plants ( Table 1). The classical anticancer effects of apigenin related to the targeting of NFκB, p53, mitogen-activated protein kinase (MAPK) and PI3 K/Akt have been widely reported [397][398][399][400]. Table 2 highlights some anticancer properties of apigenin that are less emphasized in the literature and are related to anticancer effects. The same observation is made for another flavone, luteolin, as detailed in Tables 1 and 2. Apigenin kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptotic-related cell death pathways in cancer cells, impairs the biology of CSCs, and is bioselective ( Table 2).
Naringenin is a natural flavanone found in oranges, grapefruit and tomato skin, for example ( Table 1). The antioxidant and anti-inflammatory effects of naringenin are well known [401,402]. Naringenin exerts suppressive effects on TGF-β ligand-receptor interactions [402], and TGF-β signalling controls a diverse set of cellular processes in cancer such as cell proliferation, differentiation and apoptosis [403]. Naringenin is known to display classical cytotoxic profiles on various cancer cell lines [402]. Table 2 highlights the most important effects of naringenin as an anticancer agent. Naringenin kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptotic-related cell death pathways in cancer cells, impairs the biology of CSCs, and is bioselective ( Table 2).
Genistein and daidzein are the principal soy isoflavones (Table 1). Genistein has been reported to inhibit protein tyrosine kinases, topoisomerase II, and ribosomal S6 kinase and to induce apoptosis and/or differentiation of cancer cells [404,405]. Apart from these classical effects reported for genistein and also for daidzein, Table 2 highlights the biological events that are impaired by these two food components in models of aggressive cancers. Genistein kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptosis-related cell death pathways in cancer cells, impairs the biology of CSCs, and is bioselective ( Table 2). By contrast, no report has described daidzein-induced cancer cell death via a non-apoptotic cell signalling pathway or the impairment of CSC biology ( Table 2).
Phenolic acids represent the highest polyphenol intake from food [406]. Phenolic acids are divided in two groups, i.e., hydroxybenzoic acid derivatives (including salicylic acid and gallic acid, for example) and hydroxycinnamic acid derivatives (including caffeic acid, chlorogenic acid, cinnamic acid, and ferulic acid) [406] ( Table 1). The major representative of hydroxycinnamic acids is caffeic acid, which occurs as an ester with quinic acid called chlorogenic acid (5-caffeoylquinic acid) [406]. Table 2 summarizes the most striking anticancer effects associated with various phenolic acids.
Resveratrol is a polyphenolic phytoalexin found mainly in grapes (Table 1). Classical in vitro studies have reported resveratrol-induced anticancer effects through the inhibition of COX, NF-κB and STAT-3 pathways [174,[407][408][409], resulting in apoptosis-mediated cell death in cancer cell models that are sensitive to pro-apoptotic stimuli [409][410][411]. Table 2 highlights the less conventional anticancer effects associated with resveratrol. Resveratrol kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptosis-related cell death pathways in cancer cells, impairs the biology of CSCs, and is bioselective ( Table 2).
Capsaicin is the chief pungent component in the fruits of Capsicum (Table 1). For many other food components, many publications have described capsaicin-induced pro-apoptotic effects in cancer cell lines that are sensitive to pro-apoptotic stimuli [412][413][414][415]. An original mechanism of action associated with capsaicin as an anticancer agent is related to its inhibition of the activity of the transient receptor potential vanilloid type-1 (TRPV1) channel, which is broadly distributed in the brain and in various non-neuronal tissues but also in many cancer cell types [413]. Compounds that target ion channels and/or transporters represent a promising avenue to find innovative anticancer agents [416]. Table 2 summarizes the anticancer effects mediated by capsaicin in chemoresistant cancer models. Capsaicin kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptotic-related cell death pathways in cancer cells, impairs the biology of CSCs, and is bioselective ( Table 2).
Alkaloids belong to a group of natural products with remarkable importance in drug discovery programmes [85].
Berberine is an isoquinoline alkaloid that is found in many plants ( Table 1). As already reported in the previous sections for most foodrelated components, there are numerous publications describing the inhibitory effects of berberine on the production of TNF-α, IL-6 and monocyte chemo-attractant protein 1 (MCP-1) [417] and on the expression of COX-2 and MMP-2 and MMP-9 through the MAPK and NF-κB signalling pathways [417][418][419]. Table 2 highlights some effects exerted by berberine on biological events that are directly implicated in cancer aggressiveness. Berberine kills cancer cells resistant to proapoptotic stimuli, induces non-apoptotic-related cell death pathways in cancer cells, impairs the biology of CSCs, and is bioselective (Table 2).
Piperine, a major alkaloid constituent of black pepper (Piper nigrum) ( Table 1), is known to induce cytotoxic effects in various cancer cell lines as reported in the PubMed database. Anti-NF-κB effects are also reported [420]. Table 2 illustrates some less classical anticancer effects exerted by piperine. Piperine markedly impairs the biological behaviour of CSCs (Table 2). By contrast, it is not yet known whether piperine can kill cancer cells resistant to pro-apoptotic stimuli or induce non-apoptotic cell death pathways in cancer cells (Table 2). This compound is bioselective (Table 2).
Caffeine is an alkaloid present in many plants such as cocoa beans, coffee beans, cola nuts and tea leaves (Table 1). Many publications have reported the pro-apoptotic effects of caffeine in various cancer cell lines [421][422][423]. However, Table 2 reports several publications demonstrating the anticancer effects of caffeine in chemoresistant (including apoptosis-resistant) cancer models. While it has already been demonstrated that caffeine can kill cancer cells resistant to pro-apoptotic stimuli, anti-CSC activity has not yet been reported for caffeine ( Table 2). In the same manner, it is not yet clear based on the scientific literature whether caffeine is bioselective ( Table 2).
Celastrol is a quinone methide triterpene that can be extracted from the root extract of Thunder God Vine (Tripterygium wilfordii) [95] ( Table 1). The pro-apoptotic effects of celastrol have been reported as for any other food-related components but once more using cancer cell lines that are sensitive to pro-apoptotic stimuli [424,425]. Table 2 emphasizes other anticancer effects reported for celastrol. Celastrol kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptoticrelated cell death pathways in cancer cells, and impairs the biology of CSCs, while it is bioselective (Table 2). However, it is not yet clear whether celastrol is bioselective (Table 2).
Cordycepin (3′-deoxyadenosine) is one of the "bioactive" components isolated from the medicinal mushroom Cordyceps militaris [426,427]. Cordycepin is a nucleoside analogue that is structurally similar to adenosine, except that it lacks a 3′-hydroxyl group [426]. Cordycepin is known to interfere with several biochemical and molecular processes, such as purine biosynthesis, DNA/RNA synthesis and mammalian target of rapamycin (mTOR) signalling transduction [426]. As expected, cordycepin has been shown to display pro-apoptotic effects in cancer cell lines that are sensitive to pro-apoptotic stimuli [427][428][429]. Cordycepin also exerts inhibitory effects on the NF-κB and AP-1 signalling pathways [426][427][428][429]. Table 2 shows that cordycepin also impairs biological events that are associated with cancer aggressiveness. Cordycepin kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptotic-related cell death pathways in cancer cells, and impairs the biology of CSCs, although it is not yet known whether this compound is bioselective (Table 2).
Glycyrrhetinic acid is a pentacyclic triterpenoid derivative that is present in the liquorice (Glycyrrhiza glabra) ( Table 1) and is the active aglycone of glycyrrhizin [430]. Liquorice is a well-known plant that is used to add flavour to foods, beverages, and tobacco, and it is also used as a medicinal plant. Glycyrrhizin shows poor bioavailability following oral administration. It is hydrolysed to 18-β-glycyrrhetinic acid by human intestinal microflora. Glycyrrhetinic acid naturally occurs as two diastereomer forms, 18-α and 18-β-glycyrrhetinic acid, and both diastereomers are associated with cytotoxic activity in various cancer cell lines [430]. Glycyrrhetinic acid promotes apoptosis in various cancer cell lines that are sensitive to pro-apoptotic stimuli [431]. Table 2 displays more original mechanisms of action of glycyrrhetinic acid in cancer models associated with biological aggressiveness. Glycyrrhetinic acid kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptosis-related cell death pathways in cancer cells, impairs the biology of CSCs, and is bioselective ( Table 2).
Lycopene is a highly unsaturated acyclic isomer of β-carotene and is the most abundant carotenoid in tomatoes. Other sources of lycopene are red fruits, such as rosehips, watermelons, red grapefruits, papayas, apricots and pink guavas [432] (Table 1). Lycopene displays cytotoxic effects in many cancer cell lines [432,433] with a special emphasis on prostate cancer cell lines [433] because the consumption of lycopene is inversely related to human prostate cancer [97]. Apart from the classical studies demonstrating lycopene-induced apoptosis in some cancer cell lines that are sensitive to pro-apoptotic stimuli [433,434], Table 2 highlights very important anticancer effects associated with lycopene in various models of chemoresistant cancers. Lycopene kills cancer cells resistant to pro-apoptotic stimuli, impairs the biology of CSCs, and is bioselective (Table 2). However, at present, it is not known whether lycopene can induce non-apoptotic cell death signalling in cancer cells (Table 2).
Zerumbone, a sesquiterpene extracted from the rhizome of tropical ginger Zingiber zerumbet (Table 1), is known to induce apoptosis in various cancer cell lines [435,436], possibly through the modulation of the JAK/STAT pathway [437]. Zerumbone acts on NF-κB-and COX-2regulated pathways [437]. Table 2 highlights additional anticancer effects associated with zerumbone. It has not yet been demonstrated whether zerumbone kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptotic-related cell death pathways in cancer cells, or impairs the biology of CSCs ( Table 2).
The health-related benefits of organosulphur compounds have been mainly attributed initially to the medicinal properties of garlic [438]. Garlic extracts (enriched in diallyl disulphide) are known to induce apoptosis in various cancer cell lines [439][440][441][442]. However, other organosulphur compounds also display potent anticancer activity. One example is onionin A, which is isolated from onion bulbs [438] ( Table 1). Fujiwara et al. [443] recently reviewed various mechanisms of action underlying the antitumour effects of onionin A, which include the inhibition of tumour cell proliferation, blockade of immunosuppression with myeloid cells, and suppression of cell-cell interactions between macrophages and tumour cells. Tsuboki et al. [444] further reported that onionin A suppresses ovarian cancer progression by inhibiting STAT3 signalling-induced pathways. Table 2 adds some interesting information regarding the anticancer effects contributed by onionin A. This compound is already known to kill cancer cells that are resistant to pro-apoptotic stimuli and to induce non-apoptotic cell death signalling in cancer cells (Table 2). However, it is not yet known whether onionin A impairs the biological behaviour of CSCs or is bioselective (Table 2).
Sulphoraphane is an isothiocyanate present in cruciferous vegetables (Table 1). Sulphoraphane has been reported to have multiple mechanisms of action in terms of anticancer effects, including activation of apoptosis, downregulation of antiapoptotic proteins, and antiangiogenic and antimetastatic effects [445][446][447]. Sulphoraphane also inhibits NF-κB-, HIF-1α-and Nrf2-controlled signalling pathways [447][448][449]. Table 2 provides additional information regarding the mechanisms of action exerted by sulphoraphane against aggressive cancer types. Sulphoraphane kills cancer cells resistant to pro-apoptotic stimuli, induces non-apoptotic-related cell death pathways in cancer cells, impairs the biology of CSCs, is bioselective ( Table 2).

What occurs in cancer patients when combining food-related components with chemotherapy?
Berberine inhibits the efflux of cytotoxic drugs from breast cancer cells by the MDR-related ABCG2 (the BCRP breast cancer resistance protein; [137]), thus increasing the efficacy of anticancer drugs against breast cancer cells. However, this compound activates the MDR1 (ABCB1/P-gp170) efflux pump in gastric cancer cells, consequently decreasing, for example, the efficiency of paclitaxel in gastric cancer cells [139]. Caffeic acid can also protect cancer cells against paclitaxelinduced cytotoxic effects [450] similar to daidzein and genistein [91], which can stimulate the efflux of cytotoxic agents out of cancer cells. Genistein also increases the expression levels of several MDR-related efflux pumps [451].
Several flavonoids, including apigenin (Tables 1 and 2), can inhibit CYP2C9 [452], an enzyme that is responsible for the metabolism of many drugs used to combat cancer including tamoxifen [453]. Tamoxifen is a drug widely used to treat hormone-sensitive breast cancers, and a high consumption of flavonoid-rich diets by breast cancer patients should therefore limit the efficacy of tamoxifen. Wu et al. [151] reported that breast cancer patients use alternative and natural remedies more than patients with other malignancies, with 63-83% using at least one type of alternative medicine and 25-63% using herbs and vitamins.
From a more general point of view and considering a large set of food-related components (and not only flavonoids), Haefeli and Carls [454] recently reported that 30-70% of cancer patients use complementary and alternative medicines, including herb-drug combinations, while some of these combinations (reviewed in 454) can critically alter the exposure of anti-neoplastic and palliative treatment. Izzo and Ernst [455] also emphasized that interactions between herbal medicines and synthetic drugs exist and can have serious clinical consequences; these authors stated that healthcare professionals should ask their patients about the use of herbal products and consider the possibility of herb-drug interactions. Chen et al. [456] accordingly emphasized some years ago that the clinical consequence of herb-drug interactions varies, from being well-tolerated to moderate or serious adverse reactions, or possibly life-threatening events; therefore, these authors claimed that, undoubtedly, the early and timely identification of herbdrug interactions is imperative to prevent potentially dangerous clinical outcomes. The few points that we raise here clearly indicate that cancer patients must be aware that some diets can induce considerably adverse effects against the benefits contributed by chemotherapy. Thus, we want to draw the attention of the readers to the serious problems that can appear when blindly following certain diets and anticancer therapies (and many other types of therapies). This problem occurs in industrialized countries where cancer patients have access to chemotherapeutic drugs. By contrast, this is not yet the case in many less industrialized countries. Charepalli and colleagues [112] accordingly reported that the World Health Organization (WHO) predicts a 70% increase in the cancer incidence in developing nations over the next decade and that, although these nations have limited access to novel therapeutics, they do have access to foods that contain chemopreventive bioactive compounds such as those described in the current review.

Food components and the bcl-2 family protein
Delbridge et al. [457] recently and very comprehensively reviewed the possibility of selectively targeting Bcl-2-related signalling pathways in cancer cells as new avenues to combat cancers associated with dismal prognoses. Indeed, the aberrant expression of Bcl-2 family members is strongly associated with the resistance of various cancer types to chemotherapy and radiation [458]. Thomas et al. [458] accordingly state that i) "strategies to specifically identify and inhibit critical determinants that promote therapy resistance and tumour progression represent viable approaches for developing effective cancer therapies" and that ii) "from a clinical perspective, pretreatment with novel, potent Bcl-2 inhibitors either alone or in combination with conventional therapies hold significant promise for providing beneficial clinical outcomes". Ashkenazi et al. [459] have recently reviewed the latest progress in the selective targeting of BCL-2 family proteins for cancer therapy. Many food-related components act on Bcl-2 family members, and we highlight here below some of the most important features in this specific area. Therefore, recent publications that emphasize specific actions of food-related components under review in Bcl-2-related signalling pathways are listed: • Apigenin: Chan et al. [460] and Chen et al. [ • Zerumbone: Sun et al. [481].
Thus, the current section clearly emphasizes that most of the foodrelated components under review with anticancer effects (both in vitro and in vivo) also impair the Bcl-2 signalling pathway in cancer cells.

Conclusions
We hope that the readers of the current review gain new information regarding the many food-related components that display anticancer effects. The current review highlights that some specific diets taken daily could maintain a given cancer type, even if associated with clear biological (and thus, clinical) aggressivity, in a steady state phase of growth. In other words, diets based on the consistent consumption of at least all the food-related compounds described in the current review could transform cancer as an acute disease into a chronic one, consequently increasing life expectancy. The review by Liu [482] is a major contribution in the field of "diet against cancer". Indeed, Liu [482] reported that increasing evidence suggests that the health benefits of fruits, vegetables, whole grains, and other plant foods are attributed to the synergy or interactions among the bioactive compounds and other nutrients in whole foods. Liu [482] accordingly stated that consumers should obtain their nutrients, antioxidants, bioactive compounds, and phytochemicals from a balanced diet with a wide variety of fruits, vegetables, whole grains, and other plant foods for optimal nutrition, health, and well-being, not from dietary supplements. Finally, the current review also highlights that various food-related components that display anticancer effects (such as those detailed in Table 2) cannot be combined with chemotherapy without supervision by a medical nutritionist. By contrast, an equilibrated diet, including the food-related components listed in the current review should be beneficial for cancer patients who are not undergoing chemotherapy.

Conflicts of interest
The authors declare no conflicts of interest.