Nutritional Countermeasures Targeting Reactive Oxygen Species in Cancer: From Mechanisms to Biomarkers and Clinical Evidence

A. Direct measurement of ROS production

The direct measurement of ROS produced in vivo is not really standardized yet, and a number of possible assays employ electron paramagnetic resonance (EPR) techniques and dyes or probes that become fluorescent upon reaction with ROS (88, 237, 311) (Fig. 4). Intracellular ROS can be measured in vivo either with small-molecule fluorescent probes (173, 240, 300) or with fluorescent probes based on genetically encoded constructs (237). The most common small molecule used is 2′,7′-dihydrodichlorofluorescein (H₂DCF), which is oxidized by H₂O₂ to fluorescent 2′,7′-dichlorofluorescein (DCF) (173). The small probes like DCF have the advantages of simplicity and a high signal-to-background ratio, but they are very susceptible to artificial side reactions. For example, oxidation of H₂DCF can occur also in the absence of H₂O₂ being stimulated by metals, peroxidases, and cytochrome c (237). Moreover, during the reaction, an intermediate DCF^•− radical is formed that can react with oxygen forming O₂ ^−• with the consequence of an artificial amplification of the fluorescence signal (173). Hydroethidine (HE) is the most popular fluorogenic probe to detect intracellular O₂ ^−• by formation of fluorescent 2-hydroxyethidium. However, HE was also shown to be an unreliable indicator for intracellular O₂ ^−• formation (399). Thus, there is a demand for improved chemical probes, and these are now being developed. For example, Peroxy Green 1 (PG1) and Peroxy Crimson 1 (PC1) have recently been proposed for an improved selective measurement of H₂O₂ in cells (240, 300). However, the increase in selectivity was gained at the expense of sensitivity, since the H₂O₂-mediated boronate deprotection reaction is a relatively slow and irreversible process, and therefore PG1/PC1 cannot be used for detection of transient changes in the intracellular H₂O₂ concentrations (300).

OPEN IN VIEWER

FIG. 4.

Detection of ROS by using different biomarkers. ROS can be assessed in vivo either directly or indirectly. Direct methods include fluorescence-based probes and EPR. Indirect methods include measuring products of oxidative damage, total antioxidant capacity, and measuring levels/activities of low-molecular-weight antioxidants/antioxidant enzymes. 8-oxodG, 8-oxo-7,8-dihydro-2′-deoxyguanosine; 8-oxoGua, 8-oxo-7,8-dihydroguanine; CBA, crocin bleaching assays; cpYFP, circularly permuted yellow fluorescent protein; di-Tyr, o,o′-dityrosine; EPR, electron paramagnetic resonance; FRAP, ferric ion reducing antioxidant power; GPX, glutathione peroxidases; GSH, glutathione; H₂DCF, 2′,7′-dihydrodichlorofluorescein; HE, hydroethidine; HNE, 4-hydroxy-2-nonenal; LDL, low-density lipoproteins; MDA, malondialdehyde; MS, mass spectrometry; ORAC, oxygen radical absorbance capacity; PG1, Peroxy Green 1; roGFP, reduction-oxidation sensitive green fluorescent protein; SOD, superoxide dismutase; TAPA, total antioxidant potential assay; TEAC, Trolox equivalence antioxidant capacity; TRAP, total radical trapping antioxidant parameter.

Some limitations of the chemical probes may be overcome by the development of genetically encoded redox probes that are more specific and allow quantitation of ROS production within live cells (237). The basis for these redox-sensitive probes is green and yellow fluorescent proteins (GFP and YFP). Introduction of specific cysteine residues and other changes into GFP and YFP resulted in a strong redox-dependent change in fluorescence and development of reduction–oxidation-sensitive GFPs (roGFP1 and roGFP2) and circularly permuted YFP (cpYFP) (137, 272). The roGFPs seem to be best suited to determine the steady-state redox conditions and to document redox changes that happen in the frame of minutes (237). Fusing of roGFP2 to glutaredoxin (Grx1-roGFP2) enabled development of a more sensitive probe for the measurement of glutathione (GSH) redox potential (125), whereas coupling of roGFP2 to yeast peroxidase Orp1 (Orp1-roGFP2) and cpYFP to the regulatory domain of Escherichia coli OxyR (HyPer) improved fluorescence imaging of H₂O₂ in live cells (30, 126). Although cpYFP was initially described to be also suitable for intracellular O₂ ^−• detection (375), another report claimed that cpYFP is highly responsive to pH, but not to O₂ ^−• in mitochondria (319). Thus, genetically encoded redox fluorescent probes represent a new generation of ROS-detecting probes that still require improvement, especially with respect to specficity.

Another approach to measure ROS in vivo is EPR where spin traps react selectively with free radicals to generate a relatively stable product (298, 393). In particular, spin trapping with nitrones such as 2,2,6,6-tetramethyl-1-piperidinyl-1-oxyl (TEMPO) can generate, upon reaction with ROS, more-persistent nitroxide adducts detectable by EPR (366). A 5-diethoxyphosphoryl-5-methyl-1-pyrroline-N-oxide (DEPMPO) and a 2-(2-carboxyethyl)-2-methyl-4-phenyl-2H-imidazole-1-oxide (MCPIO) spin trap can be used for the detection of •OH and O₂ ^−• by EPR and mass spectrometry (MS) (298). The latter has the advantage to detect also EPR-silent probes. Recently, a ratiometric MS approach was developed to assess mitochondrial H₂O₂ in fruit flies in vivo by using MitoB probe (72). MitoB consists of a triphenylphosphonium (TPP) cation driving its accumulation within mitochondria, conjugated to an arylboronic acid that reacts with H₂O₂ to form a phenol, MitoP. Quantifying the MitoB/MitoP ratio by liquid chromatography–tandem mass spectrometry (LC-TMS) allowed measuring the H₂O₂ levels within mitochondria (72).

B. Indirect measurement of ROS generation

Despite recent advances, direct detection of intracellular ROS is still rather problematic due to their high reactivity and instability. All the optical methods for ROS measurements, described above, are not applicable to most animal models in vivo and to humans (254). Moreover, analyzing ROS in living animals and humans by conventional EPR has significant limitations due to the use of high-frequency microwaves. Therefore, ROS generation is usually indirectly assayed by detecting specific and relatively stable biomarkers in various biological fluids (serum or plasma, urine, saliva, semen, etc.), which are convenient to obtain without invasive procedures. These biomarkers are generated due to specific reactions of ROS with certain biomolecules (268). The most commonly used indicators of high ROS levels are products of lipid peroxidation, DNA oxidation, and protein oxidation (Fig. 4).

1. Lipid peroxidation markers

Lipid hydroperoxides are nonradical intermediates of lipid peroxidation, and their measurement may provide information about ROS levels (Fig. 5A). Several chromatographic, enzymatic, and colorimetric methods have been developed for their detection (40). Since analysis of lipid hydroperoxides is quite difficult, because they are very unstable and decompose rapidly, these methods can be used only with cautious interpretation of obtained results (147). Among the most widely used membrane lipid peroxidation markers are 4-hydroxy-2-nonenal (HNE) and malondialdehyde (MDA). Both are relatively stable α,β-unsaturated aldehydes that are formed during decomposition of unstable dienic hydroperoxides; however, some α,β-unsaturated aldehydes are permanently formed under physiologic conditions during apoptosis and proliferation, and therefore their appeareance does not necessarily need to reflect oxidative stress. The usual test for MDA production is thiobarbituric acid-reactive substances (TBARS), a spectrophotometric assay that quantifies a chromogen produced by the reaction of MDA with thiobarbituric acid. Although MDA is often used as a marker for ROS-mediated lipid peroxidation, the MDA levels give usually an overestimation of ROS formation. This is due to the nutrient-derived peroxides and peroxide decomposition products that can be absorbed through the gut and therefore influence MDA levels (19). Moreover, MDA can be produced also by the degradation of endoperoxides and not only during lipid peroxidation. Thus, MDA/TBARS appears not to be a reliable test system for in vivo ROS measurements, and the European Food Safety Authority recently disapproved the use of TBARS as a biomarker for lipid peroxidation (147).

OPEN IN VIEWER

FIG. 5.

Products of oxidative lipid and DNA damage. (A) The initial products of membrane lipid peroxidation are conjugated dienic hydroperoxides. These lipid hydroperoxides decompose either into several aldehydes or, if the original fatty acid is arachidonic acid, into isoprostanes. MDA, one of the most prominent aldehyde products, forms several types of DNA adducts by attacking the nucleobases or free bases. Two major adducts are shown. (B) ROS-mediated direct DNA damage. 8-oxoGua, 8-oxo-7,8-dihydroguanine; 8-oxodG, 8-oxo-7,8-dihydro-2′-deoxyguanosine; 8-oxodA, 8-oxo-7,8-dihydro-2′-deoxyadenosine; HNE, 4-hydroxy-2-nonenal; MDA, malondialdehyde.

Low-density lipoproteins (LDL) play an important role during lipid and cholesterol metabolism. While oxidation products of cholesterol such as 7β-hydroxycholesterol may predict the development of lung cancer (210), the last decade showed that measurement of oxidized LDL (oxLDL) appears to be a good indicator for atherosclerosis and coronary artery disease (363). The oxLDL measurements with immunoassays provide a reliable and easy method, although standardization of the available oxLDL assays is still needed (147).

The F2-isoprostanes (especially 5-, 8-, 12-, 15-F2-isoprostanes), which originate from the peroxidation of arachidonic acid residues in lipids, are often considered to be the best indicators of lipid peroxidation in humans (320). They are not influenced by the diet and can be assessed in various biological fluids, including urine, plasma, synovial fluid, bronchoalveolar fluid, bile, lymph, and cerebrospinal fluid. Urine is considered most suitable for F2-isoprostane measurements, since isoprostanes in plasma have a shorter half-life, and their levels vary widely during the day. Indeed, increased levels of urinary 15-F2t-isoprostane were shown to be associated with a risk for breast cancer and hepatocellular carcinoma (309, 386). Further, the most abundant F2-isoprostane is 8-isoprostaglandin-F-2 (8-iso-PGF2), and increased 8-iso-PGF2 concentrations, as detected by gas chromatography–mass spectrometry (GC-MS), LC-MS, and enzyme immunoassays, have been associated not only with cancer but also with atherosclerosis, pulmonary diseases, and Alzheimer disease (250). However, the use of F2-isoprostanes also has its limitations. Extensive auto-oxidation of arachidonic acid occurs in blood plasma leading to artificial generation of these compounds (305). Moreover, though F2-isoprostanes are quite stable in isolated samples of body fluids, once released into circulation, they are rapidly metabolized in vivo (76). Another way of assessing whole-body lipid peroxidation is based on the analysis of exhaled hydrocarbon gases (alkanes), for example, ethane and pentane formed during decomposition of peroxyl radicals (88).

C. The limitations of oxidation markers in use

It is well known that oxidative stress-related markers, such as 8-oxodG, HNE and MDA, are commonly elevated in tumor patients and often correlate well with risk factors, such as viral infection and inflammation (223, 323, 390). For example, the correlation between multiple oxidative stress biomarkers and breast cancer, squamous cell carcinomas, leukemia, and colorectal cancer was found (106, 221, 390, 400). As discussed later in more detail, there are data showing that diet composition, especially the supplementation with antioxidants, influence the plasma or urinary levels of oxidative stress markers (see sections VI and VII) (48, 101, 232, 247). Some studies have evaluated oxidative stress biomarkers as particular cancer risk factors and simultaneously related those factors to nutrient intake. In particular, it was found that an antioxidant micronutrient cocktail can modulate oxidative stress in patients with a history of sporadic colorectal adenoma (149). In another study, the serum TAC was decreased with an increasing carbohydrate intake, but was increased with a higher vitamin A and vitamin C intake in breast cancer patients (390). It was also found that the intake of chicken and high-fat dietary products was associated with increased levels of plasma lipid peroxides, while vitamin E intake was associated with decreased lipid peroxides in women with breast cancer (364). By contrast, several other studies showed that cancer-relevant biomarkers of oxidative stress were unchanged by diet. For instance, increased consumption of selenium-rich food and subsequent increase in blood Se concentrations did not affect biomarkers of oxidative stress in healthy Australian males (387). No differences in the DNA and lipid oxidative damage markers and antioxidant capacity between intervention groups of healthy, nonsmoking men receiving 2, 5, or 8 servings/day of vegetables and fruit were found (45).

The use of biomarkers in human studies requires an ability to sensitively, specificly, and reproducibly measure oxidative damage to the relevant molecular targets (132). The substances used as biomarkers should be stable, generated only by ROS action, and not influenced by their presence in food. As it was discussed before, the existing biomarkers fulfil these demands only to some extent. It is known that different biomarkers do not measure identical aspects of oxidative stress, and while there are usually good correlations between the concentrations of several peroxidation products, including MDA, F2-isoprostanes, conjugated dienes, GSH, and PCOs, these parameters are often not correlated with other biomarkers such as the concentration of antioxidants and DNA fragmentation (88, 320). Even more, the usual ROS markers insufficiently reflect intracellular processes leading to ROS generation and are not intracellular compartment specific. Indeed, the biomarkers of lipid oxidation are generated in the plasma membrane as well as in various intracellular membranes. DNA oxidation markers cannot be considered specific to the nuclear compartment not only because of mitochondrial DNA (mtDNA) oxidation but also due to oxidation in the nucleotide pool, since single-nucleoside triphosphates are less protected from oxidative damage than the chromatin-bound genomic DNA. Although some types of oxidative products of nucleotides are not amenable to DNA synthesis (e.g., thymine glycol and 5′,8-purine cyclodeoxynucleoside), other species may be tied into the nucleic acids (322). However, none of the analytical methods used to measure DNA adducts identifies where the oxidative DNA damage is located (76).

D. Compartment-specific ROS measurement

The activity of compartment-specific antioxidant enzymes (see section V) in correlation with a disease could be very informative concerning the details of intracellular generation of ROS. A good example of such enzyme that has compartment-specific isoforms is SOD, and changes in the cytosolic Cu/ZnSOD (SOD1), mitochondrial MnSOD (SOD2), or extracellular SOD (EcSOD; SOD3) activity or expression can be correlated to ROS levels. Elevated MnSOD levels were found in tissues of patients with gastric carcinomas and adenomas (182) as well as squamous cell carcinomas, whereas no significant changes in Cu/ZnSOD were detected (106, 391). By contrast, increased Cu/ZnSOD activity was found in tissue samples from patients with colorectal cancer (331), in leukemia patients (400), and in patients with breast cancer (22, 288). EcSOD (but not Cu/ZnSOD or MnSOD) expression was found to be significantly decreased in lung cancer (350). However, the obvious problem in such studies is the limited availability of healthy human tissues in which antioxidant enzyme activities can be assayed. Although the changes in antioxidant enzymes may correlate the ROS levels to intracellular compartments, the significance of these changes is often difficult to interpret, since expression of antioxidant enzymes is often induced by oxidative stress itself, and, moreover, their decreasing activities and levels upon oxidative stress, probably due to structural modification by free radicals, were also reported (67, 230). For instance, the levels of catalase, a rather peroxisome-specific antioxidant enzyme, were found to be induced (288), decreased (331), and unaffected (22) in different types of cancer. In line, while high MnSOD expression correlates with a poor prognosis, advanced progression, and an invasive and metastatic phenotype of gastrointestinal tumors (182), it is a factor correlated with favorable survival in patients with buccal mucosal squamous cell carcinomas (106).

While the biological significance of changes in antioxidant enzymes is under discussion and a common view has not been reached, methods for quantitative detection of compartment-specific ROS generation in live cells is needed. Recently, a new class of compartment-specific reporters for H₂O₂ imaging was generated (300, 337). Thereby, a mutant of the DNA repair protein O6-alkylguanine-DNA alkyltransferase (SNAP tag) was targeted to the different compartments and later conjugated with PG1. The SNAP-PG adducts respond to H₂O₂ with an increase of green fluorescence in the target organelle. However, these probes react irreversibly with H₂O₂, and cannot be used to monitor transient changes in H₂O₂ concentrations. By contrast, genetically encoded ROS probes such as roGFPs (Fig. 6) and HyPer can be targeted to distinct subcellular compartments and react in a reversable manner.

OPEN IN VIEWER

FIG. 6.

Measuring real-time redox changes in living cells. (A) roGFPs were expressed in different cellular compartments of HeLa cells and imaged by live-cell microscopy. Cells were treated with dithiothreitol (DTT) as indicated. After backbone subtraction, the ratio of the fluorescence intensity between 420 nm and 470 nm was plotted against the time. The curve represents the mean of three independent measurements. Vertical lines indicate the time point at which the compartments started to change the redox state after addition of DTT. (B) Representative images of cells expressing redox-sensitive GFP in the Golgi, endoplasmic reticulum (ER), and mitochondria (mito). 4′,6-diamidino-2-phenylindole (DAPI) staining indicates the nucleus.

Another important progress was made with generation of mitochondrion-targeted probes for ROS detection (Fig. 7) (82). The promising strategy for fluorescence imaging of mitochondrial ROS in living cells is to create dyes that contain both a ROS-responsive element and a mitochondrion-targeting moiety. Mitochondria Peroxy Yellow 1 (MitoPY1) is an example of such a dye (81). Reaction of MitoPY1 with H₂O₂ triggers a fluorescence increase by its conversion to MitoPY1ox. Mitochondrial O₂ ^−• may be assayed by using mitochondrion-targeted HE (Mito-HE, MitoSOX red), a TPP⁺ conjugated to HE via an alkyl chain. Further, a recently developed mitochondrion-targeted fluorescent lipid peroxidation probe MitoPerOx was shown to report mitochondrial lipid peroxidation in isolated liver cells (284). Highly reactive ROS, including •OH, peroxynitrite (ONOO⁻), and hypochlorous acid (HOCl), can be measured in mitochondria by rhodamine-like fluorophores capped with either a 4-aminophenyl aryl ether (MitoAR) or a 4-hydroxy aryl ether (MitoHR) group (187). However, all the limitations attributed to the small fluorescent probes discussed before (see section III.A) are also applicable to mitochondrion-specific small probes (237, 254, 399). Alternatively, genetically encoded redox probes such as roGFPs and Grx1–roGFP2 were successfully targeted to mitochondria in several cell lines in vitro (237) and in vivo in transgenic mice (128). Those analyses could corroborate that mitochondria are most reduced organelles, whereas the ER and Golgi are much more oxidized. Accordingly, the more oxidized compartments appear to react faster when cells are challenged by a reducing agent or environment (Fig. 6). Further, the use of genetically encoded redox probes and a new chemical strategy to trap the in vivo redox state of the transgenic biosensor enabled quantitative mapping of the GSH redox potential and H₂O₂ in the cytosol and mitochondria during development and aging in Drosophila (9).

OPEN IN VIEWER

FIG. 7.

Approaches for mitochondrion-specific ROS detection. Fluorescent probes, either small molecule-containing (mitoPY1, mito-HE, and mitoHR) or genetically encoded (roGFP and HyPer), are underlined. The genetically encoded fluorescent probes can be investigated by live-cell microscopy. The generation of Mito-DEPMPO-superoxide adducts can be measured by EPR. The MitoB/MitoP ratio can be assayed by LC-TMS. 2-OH-E⁺, 2-hydroxyethidium; DEPMPO, 5-diethoxyphosphoryl-5-methyl-1-pyrroline-N-oxide; EPR, electron paramagnetic resonance; HE, hydroethidine; HMTMR, hydroxymethyl tetramethylrhodamine; LC-TMS, liquid chromatography–tandem mass spectrometry; mito, mitochondria; MnSOD, mitochondrial superoxide dismutase; PY1, Peroxy Yellow 1; roGFP, reduction-oxidation sensitive green fluorescent protein.

Despite recent progress, it should be noted that the optical approaches to compartment-specific ROS measurement are often applicable only to cultured cells or to optically amenable organisms such as zebrafish or worms. Two-photon confocal microscopy technique is needed to probe deeper beneath the surface in nontransparent transgenic animals expressing fluorescent proteins (254). As an alternative to optical ROS detection, the use of mitochondrion-targeted nitrone traps for EPR detection of ROS (Mito-DEPMPO) was proposed (139) as well as the ratiometric MS probe targeted to the mitochondrial matrix (MitoB), which showed an increase in mitochondrial H₂O₂ production with age in Drosophila (72).

The potential for the development of markers indicating high ROS levels in a compartment-specific manner was confirmed by the effects of antioxidants designed to selectively remove ROS from mitochondria in vivo (53). The most extensively studied mitochondrion-targeted antioxidant is mitoquinone (MitoQ); it contains an antioxidant quinone moiety that is covalently linked to a lipophilic TPP⁺ via a 10-carbon alkyl chain. The TPP⁺ part allows that MitoQ efficiently accumulates in mitochondria in a membrane potential-dependent manner (333). MitoTempol and MitoVitE are other promising mitochondrion-targeted antioxidants derived from tempol (and antioxidant nitroxide) and vitamin E, respectively, covalently attached to TPP⁺. Mitochondrion-targeted peptides and plastoquinone derivatives have been developed as well (332, 343). Mitochondrion-targeted antioxidants were used in a number of in vivo studies in rats and mice and found to be protective against oxidative damage in many different pathologies, such as diabetes, cardiovascular disease, cancer, and neurodegenerative disorders (61, 122, 217). In line, MitoQ oral administration decreased protein oxidative damage markers as well as mitochondrial markers of oxidative damage in several animal models (342), though it did not influence expression and activity of mitochondrial MnSOD in wild-type mice (306). It was recently shown that mitochondrion-targeted antioxidants could protect pancreatic β-cells against oxidative stress under glucotoxic and glucolipotoxic conditions (207). However, the results of two phase II human trials (with Parkinson's disease and chronic hepatitis C patients) are less convincing (333). In this respect, it should be taken into account that presence of high quantities of TPP⁺-conjugated probes in the mitochondria may lead to quenching, mitochondrial uncoupling, and changes in the mitochondrial function (53, 165).

A. ROS-induced oxidative damage

Different ROS have an unequal ability to damage DNA and proteins. While •OH are known to cause most damage to DNA by reacting with all four bases, ¹O₂ selectively attacks guanine (381). It is unlikely that O₂ ^−• and H₂O₂ react with DNA bases directly, though they contribute to DNA damage by forming •OH. The ROS mostly responsible for protein damage is •OH that acts by abstracting a hydrogen atom from the protein polypeptide backbone (359). The abstraction of a hydrogen atom leads to the formation of a carbon-centered radical initiating the formation of various products such as peroxyl radicals, alkyl peroxide, and alkoxyl radicals (338). In addition, lipid peroxidation contributes to the formation of DNA adducts (see section III.B.2, Fig. 5A) as well as etheno-adducts and propano-adducts, which are also considered to be genotoxic (258, 359).

In line, a number of reports indicate that nuclear DNA damage appears to be important in tumorigenesis. In comparison to healthy individuals, an almost 50% higher level of the DNA oxidation product 8-oxoGua has been detected in the lung, breast, and prostate cancer patients and has been implicated to cause the Ras and p53 mutations found in both lung and liver cancer cells (352). In comparison to its nuclear counterpart, mtDNA is even more vulnerable, since it lacks histones, has limited repair capacity, and lies in a closer proximity to the sites of mitochondrial ROS generation (54). In line, mtDNA mutations have been observed in a great number of human cancers (54). Although mutation of almost all genes within the mitochondrial genome is affiliated with various types of tumors (244), one particular region in the mtDNA, which is called displacement loop (D-loop), is known as a mutational hotspot. The D-loop mutation is associated with hepatocellular carcinoma (348), ovarian cancer (214), breast cancer (349), colorectal cancer (206), and melanoma (347).

Epigenetic changes such as DNA methylation can also be influenced by ROS (102). For example, ROS-mediated DNA base modifications such as 8-oxodG formation adjacent to a cytosine may prevent methylation of the latter (353). The resulting hypomethylation may contribute to induction and suppression of gene expression, thus enabling the cells to achieve novel characteristics that may contribute to tumorigenesis. ROS are also associated with hypermethylation (102) that can lead to increased overall mutation rates by downregulating expression of repair genes such as O-6-methylguanine-DNA methyltransferase (MGMT) and mutL homolog 1 (MLH1) (398).

B. ROS-dependent regulation of signaling pathways

The involvement of ROS in the carcinogenesis is also connected to the physiological role of ROS in several major signaling cascades (175, 249, 371). Interestingly, growth factors, hormones, and coagulation factors, including platelet-derived growth factor (PDGF), angiotensin II, and insulin, have been shown either to lead to the generation of ROS or to use them at least partially as messengers. Therefore, an aberrant accumulation of ROS within the cell may imitate the signaling function of these stimuli and lead to a ligand-independent modulation of these signaling cascades. Although not all of the ROS-mediated interventions are tumorigenic, a majority of them are supposed to be. This appears often to be due to stimulation of various components of the mitogen-activated protein kinases (MAPK) and/or phosphatidylinositol 3-kinases (PI3K)/Akt signaling cascades, while a number of other pathways, including protein kinase C (PKC), wnt/β-catenin, hedgehog, and Notch signaling, can be also modulated by ROS (73, 197, 371, 389). Since intracellular pathways can be influenced by ROS in a bewildering number of ways (294), we will provide here only some typical examples that are also presented in Figure 9.

OPEN IN VIEWER

FIG. 9.

ROS-regulated signaling pathways. Schematic and simplified presentation of major intracellular signaling pathways that can be influenced by ROS either positively or negatively. Relatively low concentrations of ROS necessary for regulation of signaling pathways described in the figure are mostly attained by specific enzymatic reactions as well as due to the changes in cellular metabolic activity, leading to altered ROS production in mitochondria (see section II). DAG, diacylglycerol; ERK, extracellular signal-regulated kinase; FAK, focal adhesion kinase; GP, G-protein; GPCR, G-protein-coupled receptor; Grb2, growth factor receptor-bound protein 2; HIF-1α, hypoxia-inducible factor-1α; IκB, inhibitor of NF-κB; IKK, IκB kinase; MAPK, mitogen-activated protein kinases; MEK, MAPK/ERK kinase; MKP3, mitogen-activated protein (MAP) kinase phosphatase/dual specificity protein phosphatase-6; PHD, prolyl hydroxylase; PI3K, phosphatidylinositol 3-kinases; PI3P, phosphatidylinositol 3-phosphate; PKB/Akt, protein kinase B; PKC, protein kinase C; PTEN, phosphatase and tensin homologue deleted on chromosome 10; Raf, ras attachment factor; Ras, Rat sarcoma; RTK, receptor tyrosine kinase; SOS, son of sevenless; Shc, SHC-transforming protein; Src, sarcoma.

C. Dual role of ROS in tumorigenesis

Together, the mechanisms for redox regulation of the major intracellular signaling pathways are complex and not yet resolved until the last detail, thus requiring further investigations. Since the function and regulation of the major intracellular cascades such as MAPK and PKB were comprehensively reviewed (57, 289, 294), the present review will not describe how each of the pathways is involved into tumorigenesis. In brief, depending on the kinetics of activation and inactivation, the subcellular localization, the formation of complexes with partner proteins, and the availability of substrates, these pathways largely contribute to cell cycle progression, apoptosis and survival, angiogenesis, invasion, and metastasis (124).

The general picture of ROS involvement into malignant transformation is further complicated by the fact that while moderately increased ROS levels lead to tumor-promoting mutations of oncogenes and tumor suppressor genes as well as to induction of signaling pathways responsible for carcinogenesis, excessive ROS levels can suppress tumors (189). A number of studies demonstrated the ability of ROS to induce apoptosis of cancer cells; some studies also found that high ROS levels correlate with decreased proliferation of transformed cells, decreased tumor angiogenesis, and metastasis [listed in (124)]. A threshold concept for cancer cell therapy implies that it should be possible to preferentially accumulate ROS in cancer cells above the cellular tolerability level, thus inducing their death (189, 277). In accordance with this concept, a recent study demonstrated that piperlongumine, an alkaloid from Piper longum, increases ROS levels and apoptotic cell death in both cancer cells and normal cells engineered to have a cancer genotype, but it has little effect on either rapidly or slowly dividing primary normal cells (287). Therapeutic strategies exploiting the role of ROS both as an inducer and an inhibitor of tumorigenesis are being developed (294, 351). For example, procarbazine, motexafin gadolinium, elesclomol, 2-methoxyestradiol, and imexon are used to increase ROS, while minodronate and histamine are used to eliminate ROS (124). Similar to many chemotherapy drugs, radiation therapy also stimulates ROS generation, thus eliminating cancer cells (20).

A. Intracellular and extracellular localization of antioxidants

Various components of the antioxidant system are unequally distributed between different intracellular compartments (Fig. 12 and Table 2). For example, in eukaryotic cells, the Cu- and Zn-containing Cu/ZnSOD is located mainly in the cytosol and peroxisomes, whereas another Cu-, Zn-dependent form of SOD, known as EcSOD, is secreted and forms a glycosylated heterotetramer bound to the extracellular matrix (141). In contrast, an Mn-dependent MnSOD is found in the mitochondrial matrix.

OPEN IN VIEWER

FIG. 12.

Intracellular localization of major reactions that neutralize ROS. Different antioxidant enzymes and their isoforms are localized in various intracellular compartments or extracellularly, thus leading to unequal distribution of generation/detoxification sites for major ROS inside the cell. When O₂ accepts an electron, superoxide (O₂ ^−•) is formed. Although spontaneous dismutation to H₂O₂ may occur, this process is usually catalyzed by SOD. Since H₂O₂ gives rise to the formation of the highly reactive •OH, it is usually detoxified in cells by the action of GPX, CAT, and Prx. Extracellular ROS can be neutralized by specific isoforms of SOD and GPX. CAT, catalase; ec, extracellular; ER, endoplasmic reticulum; GPX, glutathione peroxidases; GR, glutathione reductase; GSH, glutathione; GSSG, glutathione disulfide; NOX, NADPH oxidase subunit; Prx, peroxiredoxins; SOD, superoxide dismutase.

GPX1 is the ubiquitous enzyme that reduces cytosolic, mitochondrial, and, in some cells, peroxisomal peroxides, while very similar GPX2 is an epithelial-specific form with highest expression in the intestine (219). A secreted GPX3 form expressed in the lung and kidney and measurable in plasma can be important in the protection against external peroxides (219). There are three different isoforms of GPX4 (also known as phospholipid–hydroperoxide GPX) derived from a single gene: cytosolic (c-GPX4), mitochondrial (m-GPX4), and nuclear (n-GPX4) (155, 266). Unlike other GPX, GPX4 has a broad range of substrates; in addition to H₂O₂, these include derivatives from cholesterol and cholesteryl esters and thymine hydroperoxide (155).

Prx1, 2, and 4 function mainly in the cytoplasm, but are also expressed in the nuclei; Prx1 is also present in the mitochondria and peroxisomes, whereas Prx4 is found in lysosomes (116, 157, 267). While Prx3 is localized in the mitochondria (267), Prx5 is found in the mitochondria, cytoplasm, nuclei, and peroxisomes (324). Similarly, Trx1 as well as TrxR1 can be found in the cytoplasm and nucleus. By contrast, Trx2 and TrxR2 are localized to mitochondria, where they appear to have a protective function against oxidative stress (116, 267).

In addition to antioxidant enzymes, several mitochondrial electron carriers such as cytochrome c and ubiquinol possess antioxidant activity. Interestingly, the inner mitochondrial membrane appears to contain a high amount of the antioxidant vitamin E (134), whereas GSH and vitamin C are important antioxidants in the mitochondrial matrix (53). Lysosomes contain high concentrations of low-molecular-weight antioxidants such as GSH and vitamin C (195).

Data concerning oxidative stress induced by depletion of glutamate and glutamine indicate that the nuclear compartment may have a better protection against oxidative stress than the cytoplasm or mitochondria (117). Indeed, the nucleus contains antioxidants such as GSH, n-GPX4, Trx1, TrxR1, and Prx1, 2, 4, and 5. Data from Prx1-deficient cells show a higher DNA damage (93), and together with nuclear expression of GSTpi in cancer tissues associated with anticancer drug resistance indicate that the regulation and functions of the nuclear antioxidant system are distinct from similar processes in the cytoplasm (116).

B. The functions of antioxidant enzymes unrelated to their ROS-removing activities

Beside their activities in removing excessive ROS described above, antioxidant enzymes are involved in a wide range of cellular processes. The balance between ROS generation and removal is necessary to keep physiological processes running. The floodgate model proposes that while low fluxes of H₂O₂ are scavenged by antioxidant enzymes, primarily Prx, higher fluxes cause localized Prx inactivation, enabling redox-sensitive signaling proteins such as PTPs to be targeted (384). It was recently found that Prx1 associated with the membranes is transiently phosphorylated on Tyr194, and thereby inactivated both in cells stimulated by growth factors and in those at the margin of healing cutaneous wounds in mice (382). Thus, the localized inactivation of Prx1 allows the transient accumulation of H₂O₂ around the membranes, where signaling components are concentrated, while preventing the toxic accumulation of H₂O₂ elsewhere. In addition, GPX4 was found to be involved in an apoptosis-inducing factor (AIF)-mediated cell death pathway (321) and to be pivotal for male fertility (356). As another example, the Trx/TrxR system supports the activity of ribonucleotide reductase, inhibits apoptosis signal-regulated kinase-1, and controls a pathway for methionine sulfoxide reductase B2 reduction mediated by metallothionein (116).

Although excessive ROS generation is still considered one of the causes for malignant transformation of normal cells, this does not necessarily mean that antioxidants have antitumorigenic properties. While almost any chemical reducing an oxidant's activity in vitro could be classified as an antioxidant, this may occur independent of its ability to have beneficial effects in vivo (127). In addition, some molecules such as vitamin C and plant polyphenols, including quercetin, which are considered to be antitumorigenic, have an ability to act both as an antioxidant or a prooxidant depending on their concentration (see section VIII.A) (130, 296, 362). In fact this again seems to be counterintuitive, many antitumor agents act by inducing ROS production in cancer cells (see section IV.C). One of the possible ways of how elevated ROS inhibit cell proliferation is p16-dependent sustained activation of PKC δ (290, 346). In line, antioxidants not only protect normal cells but also can favor tumor growth. In particular, they were found to promote survival of cells that are metabolically impaired and therefore lack attachment to the extracellular matrix (315).

Together, it becomes clear that ROS are required for normal cellular functions while its excessive generation leads to development of diseases, including cancer. Certain threshold concentrations of ROS exist at which either survival or cell death responses may be triggered; these may differ dependent on the cell/tissue type and/or the cellular compartment in which ROS are generated.

A. Vitamin E

Vitamin E is a term for a family of structurally related lipid-soluble components from which the naturally occurring members, the tocopherols and tocotrienols, are physiologically relevant with respect to ROS. They possess an ability to quench ROS in cell membranes and other lipid environments, thereby protecting against oxidation of polyunsaturated fatty acids (PUFA) (Fig. 13A). The best-known form of vitamin E is α-tocopherol, which is highly active in neutralizing peroxyl radicals. The reaction of peroxyl radicals with α-tocopherol is 10,000 times faster than with PUFA (49). During interaction with peroxyl radicals, α-tocopherol donates a phenolic hydrogen atom and generates an α-tocopheroxyl radical that is relatively stable (41). Under normal conditions, the α-tocopheroxyl radical does not react with the membrane components. Rather, it reacts with a second radical to form a nonradical product, or it can be reconverted to α-tocopherol by vitamin C or other water-soluble reductants (41).

OPEN IN VIEWER

FIG. 13.

Neutralization of ROS. Reactions involved in neutralization of ROS by action of (A) α-tocopherol, (B) ascorbate, (C) β-carotene, and (D) resveratrol. Tocopherol and ascorbate react not only with ROS but also with other radicals, especially with lipid radicals and lipid peroxyl radicals.

The protective role of vitamin E against ROS was indicated in several studies. By using a TBARS assay as a marker for ROS-mediated MDA production, it was found that dietary supplementation of vitamin E reduced the plasma or urinary TBARS levels (48, 232, 247). However, a number of other studies did not show the protective effect of vitamin E against oxidative damage in human subjects (261), and it should be considered that results from TBARS assays, as discussed in section III.B1, are no longer considered reliable to determine oxidative damage. In line, it has been argued that at physiological concentrations, α-tocopherol does not act as an antioxidant in vivo (261).

In general, vitamin E has been found to decrease malignant transformation in cell cultures and to suppress tumor development in animal models (Table 3) (6, 27, 41, 340). Tocopheryl phosphate, a more-active, natural derivative of α-tocopherol, was shown to have even more potent effects than α-tocopherol itself in terms of inhibition of cell proliferation and regulation of gene expression in vitro (261). It should be noted however that the tocopherols (α, β, γ, and δ-tocopherol) can decrease cell proliferation and inhibit tumor progression not necessarily due to their antioxidative properties, but via inhibition of several protein kinases, reduction of vascular endothelial growth factor production, decreased inflammation, etc. (18, 261, 299).

Epidemiological studies have been unable to point out a strong link between tumor suppression and vitamin E supplementation. For example, while one nested case–control study carried out in the United States (prostate, lung, colorectal, and ovarian cancer screening trial) found a significant inverse association between serum α-tocopherol and prostate cancer risk, primarily among smokers (377), another case–control study (prostate testing for cancer and treatment, UK) found no evidence that vitamin E was associated with the overall prostate cancer risk or more aggressive prostate cancer phenotypes (113). Moreover, neither dietary nor supplemental intakes of nutrients often suggested for prostate cancer prevention, including vitamin E and selenium, were significantly associated with prostate cancer risk in the prospective cohort study Prostate Cancer Prevention Trial (PCPT; USA and Canada, 1994–2003) (192). Regular vitamin E supplement use for more than 10 years in a particularly large prospective cohort Cancer Prevention Study II involving 991,522 U.S. adults was associated with a reduced risk of bladder cancer mortality, but regular intake for a shorter duration was not protective (163).

The Linxian large-scale randomized intervention trial, carried out in China, involved thousands of malnourished people showing that α-tocopherol was potentially associated with reduction of stomach cancer when supplemented together with β-carotene and selenium (286). In total contrast, the α-tocopherol–β-carotene study (ATBC) found that α-tocopherol is associated with a higher frequency of stomach cancer while being protective against prostate and colorectal cancer; no effect was seen on lung and pancreatic cancer (8, 293). The beneficial and adverse effects of supplemental α-tocopherol in ATBC disappeared during postintervention follow-up (369). Low-dose supplementation of various antioxidants, including vitamin E, lowered total cancer incidence and all-cause mortality in men in a randomized, placebo-controlled Supplémentation en Vitamines et Minéraux Anti-oxydants (SU.VI.MAX) trial among French adults (140). A number of other randomized intervention trials with vitamin E, either alone or in combination, did not influence or prevent cancer development (119, 211).

B. Vitamin C

Vitamin C/l-ascorbic acid is a nutrient with a variety of functions. The majority of mammals can synthesize it from glucose, whereas humans and a few other mammalian species lost the gene encoding l-gulonolactone oxidase, which is necessary for de novo synthesis of vitamin C (32). Therefore, in vivo data obtained from vitamin C supplementation in rodents cannot be simply transferred to humans. Vitamin C is an essential component of the human diet. By acting as a nucleophile, ascorbic acid can neutralize several types of ROS (Fig. 13B). During this process, it can donate an electron, becoming an ascorbyl radical (semihydroascorbate), which might be followed by a second donation of an electron that would give rise to the formation of dehydroascorbate (130). The relative unreactivity of the ascorbyl radical makes vitamin C an effective antioxidant. Apart from neutralizing ROS directly, ascorbic acid plays a critical role in revitalizing other antioxidant nutrients, for example, vitamin E, and thus along with GSH, vitamin C acts as a major redox hub (100). In line with its ROS-counteracting functions, vitamin C has also been found to modulate ROS actions in signaling pathways, such as NF-κB, c-Fos, and c-Jun (234).

It was reported that individuals suffering from severe ascorbic acid shortage show elevated levels of 8-oxodG, a prominent product of DNA oxidation (101). Ascorbic acid supplementation was also reported to have a protective role on peroxidation of plasma lipoproteins induced by acute smoking (138), while another study in which ascorbic acid was given to individual smokers failed to decrease urinary excretion of 8-oxodG (283). In line, the reduction of ROS and the inhibitory role of ascorbic acid in tumorigenesis have been a controversial issue for over two decades (36). Although several early studies suggested that ascorbic acid could provide protective effects against tumorigenesis (36, 52), later studies suggested that such an antitumor role of ascorbic acid is hitherto inconclusive (Table 3) (50, 145, 163).

C. Carotenoids

Carotenoids are a large group of tetraterpenoid organic pigments that naturally occur in various photosynthetic organisms. Although they can readily react with ROS such as peroxyl radicals or alkoxyl radicals donating electrons, their principle role seems to be ¹O₂ quenching (234). One carotenoid molecule is supposed to quench about 1000 ¹O₂ molecules (Fig. 13C) (191). Among over 700 carotenoids identified so far, ∼40 are normally found in human diet, but not all of them are uniformly retained in our body. Actually <20 carotenoids and related metabolites were detected in human tissues and plasma (264). The prominent members are α-carotene, β-carotene, lycopene, lutein, and β-cryptoxanthin. Although their role as antioxidants is well evident in vitro, the results obtained in vivo are less convincing (191, 232).

Probably due to its additional role as a precursor of vitamin A, the most studied carotenoid is β-carotene. It is speculated to have antitumorigenic properties, but the results of epidemiological studies are inconclusive, as some support its protective role, (80, 159, 170) while others do not (Table 3) (177, 312). The protective effect of β-carotene was tested in combination with α-tocopherol in the ATBC study, and after eight years of dietary supplementation, no protective effect against cancer was found (8, 293). It was even concluded that such supplementation might be potentially harmful, because β-carotene was found to be associated with increased cancer, most notably lung, prostate, and stomach, in the ATBC study (8, 369), as well as higher lung cancer mortality in the β-carotene and retinol efficacy trial (CARET) (118). In general, a recent meta-analysis of randomized controlled trials showed that β-carotene supplementation has no beneficial effect on cancer prevention, but it was associated with an increased risk not only of lung cancer but also of gastric cancer at doses of 20–30 mg/day, in smokers and asbestos workers (89), as well as significantly increased the risk of urothelial cancer (167).

By contrast, α-carotene has been reported to show more protection against skin, lung, liver, and colon cancer than β-carotene in murine models (256, 264). Apart from α- and β-carotene, several other carotenoids such as lycopene, lutein, and β-cryptoxanthin have also been reported to be potentially tumor suppressing (Table 3) (264, 327). However, the absence of large-scale randomized placebo-controlled human trials precludes us from concluding about the antioxidant and antitumorigenic property of these carotenoids.

D. Polyphenols

Dietary polyphenols are the most abundant antioxidants in human diets (136). There are almost 8000 structural variants of polyphenols ranging from single-ring-containing simple phenolic acid to highly polymerized condensed tannins, naturally present in plants as secondary metabolites. In principle, the aromatic rings of polyphenols that bear one or more hydroxyl moieties enable them to act as ROS scavengers and metal chelators. A large number of in vitro studies suggest that polyphenols are highly efficient antioxidants, since the reaction rate constants of several flavonoids with ROS in vitro surpass those of vitamin E (91).

Resveratrol is a polyphenol that has been drawing high interest of nutritionists for decades by being a candidate that may resolve the so-called French paradox (184). The role of resveratrol in diminishing oxidative stress and tumor formation was reported by several studies (Table 3 and Fig. 13D) (103, 202, 317). Pretreatment of mouse skin with resveratrol restored control levels of H₂O₂ and GSH as well as myeloperoxidase, GR, and SOD activities in the model of 12-O-tetradecanoylphorbol-13-acetate (TPA)-promoted mouse skin tumors (166). Resveratrol exerted a strong inhibitory effect on O₂ ^−• and H₂O₂ production in macrophages stimulated with lipopolysaccharides or phorbol esters (224). In addition, resveratrol can modulate signal transduction and gene expression by nonantioxidant mechanisms (299). For example, resveratrol was shown to exhibit antitumor properties by downregulating several cellular signaling pathways such as ERK1/2, Akt, and NF-κB, and by activating the deacetylase sirtuin 1 (194, 299). Although resveratrol appears to provide a significant protection against ROS-mediated tumorigenesis, the low bioavailability does not contribute to its wide use (194, 260). Moreover, there are data indicating that resveratrol at low concentrations elicits prooxidant effects and inhibits induction of apoptosis by anticancer drugs such as vincristine and daunorubicin in human leukemia cells (4, 5). However, the promises of resveratrol are apparent warranting large-scale clinical trials that could provide a more concrete knowledge about the effect of resveratrol as an antitumorigenic nutrient.

Quercetin is another prominent member of polyphenols, and some murine studies pointed to a protective role of quercetin; however, several other studies failed to find its protective effect on ROS scavenging in vivo (136, 362). Several studies identified the role of apigenin as an antioxidant and antitumorigenic nutrient (276), and to our knowledge, one ongoing placebo-controlled intervention trial has been undertaken that involves apigenin together with epigallocatechin (119). Protective roles of other dietary polyphenols (curcumin, catechins, isoflavones, anthocyanidins, etc.) against carcinogenesis in various cell culture models were also reported, and their potential role as an antitumorigenic nutrient seems to be a busy field of research (Table 3) (28, 344, 388).

Taken together, dietary polyphenols exhibit a promising role as efficient ROS scavengers, and thereby considered to be potential antitumorigenic nutrients. Although their antioxidant capacity was shown experimentally, some features obviate us from being confident that they act only by protection against ROS-mediated tumorigenesis. Recent studies also suggest that many polyphenols exert their activity indirectly by activating the transcription factor Nrf2 (314). Under basal conditions, Nrf2 is bound with the cytoskeletal anchor Kelch-like ECH-associated protein 1 (Keap1), which targets Nrf2 for ubiquitination and proteasomal degradation (345). Several stimuli, including oxidative stress and nutritional polyphenols such as curcumin and catechins, induce dissociation of the Nrf2-Keap1 complex and nuclear localization of Nrf2, where it activates transcription at the antioxidant response element (ARE) (397). The ARE controls expression of genes involved in antioxidant defense such as GST A1 and A2, heme oxygenase 1, and NAD(P)H:quinone oxidoreductase (314, 397). In line, rats consuming polyphenol-rich apple juice as compared to polyphenol-free control juice showed induced expression of Nrf2 as well as catalase, GPX2, and GR (335). However, induction of Nrf2 needs to be interpreted with caution, since several tumors display enhanced levels of Nrf2 (345). In addition, a recent study showed that Nrf2 contributes to a proliferative phenotype by favoring glutamine metabolism and nucleotide synthesis (242).

The action of polyphenols is also hampered by rapid metabolism in human tissue and by gut bacteria. This may result in very low bioavailability of polyphenols, which could be insufficient to exert any significant antioxidant activity (313). Many metabolic products of polyphenols (e.g., methylated or glucuronated forms) are expected to have reduced or even no antioxidant activity due to the loss of radical scavenging hydroxyl groups (133). In line, a comprehensive recent analysis found no convincing evidence to show that antioxidant effects of polyphenols are relevant for cardiovascular health in humans (147). Nevertheless, protective effects of polyphenols are often observed, though it is suggested that such roles might be substantially localized to the gastrointestinal tract that usually remains glutted with various oxidants (133). In general, the lack of randomized controlled trials with isolated polyphenols using biologically and methodologically valid markers makes it impossible to conclusively evaluate their antitumorigenic potential.

E. Selenium

Selenium is a trace element that plays essential biological roles (295). While excessive intake of Se leads to toxicity known as selenosis, trace amounts of Se are necessary for the proper action of more than 20 selenoproteins (29, 193). Most selenoproteins are enzymes with an active site containing Se in the form of selenocysteine. The importance of Se in the antioxidant system comes from the fact that selenocysteine is present in the catalytic sites from various members of the GPX and TrxR families. Selenoprotein P is also implicated to have antioxidative properties (295). Indeed, Se-containing compounds exerted an ROS-scavenging effect in vitro and in various animal studies (64, 95, 201). Thus, GPX and TrxR were assumed to have protective effects against carcinogenesis (43, 56, 295).

A significant inverse relationship between geographical availability of Se and carcinogenesis has been observed (69), and a similar inverse association has also been found between baseline plasma Se levels and the risk of advanced prostate cancer in many, but not all case–control and prospective cohort studies (Table 3) (11, 152, 192, 204, 279). The antitumorigenic effect of Se and selenoproteins for prostate cancer was then further supported in a mouse model (86) as well as in a study showing that the GPX1 polymorphism can play an important role in predisposition to prostate carcinogenesis (16). Similarly, GPX2 has been reported to inhibit carcinogenesis by exerting hydroperoxide removal and suppression of COX2 expression (26, 44). In line, GPX1 knockout mice are highly sensitive to oxidative stress (78), and GPX2 knockout mice were found to be more susceptible to skin cancer upon γ-irradiation (372).

These data are contrasted by a multicenter, double-blind, randomized, placebo-controlled Nutritional Prevention of Cancer Trial (NPCT), showing that dietary supplementation of Se was associated with a significantly elevated risk for squamous cell carcinoma and nonmelanoma skin cancer (90). However, in the same study, the results from secondary endpoint analyses indicated that Se supplementation is correlated with reduced incidence of lung, colon, and prostate carcinomas (70). This result was in conformity with the Linxian and SU.VI.MAX trials (140, 222, 376). A NPCT substudy found that a 200 μg/day Se treatment decreased cancer by 25%, whereas 400 μg/day of Se had no effect (297). Moreover, the selenium and vitamin E cancer prevention trial (SELECT) showed no protective effects of Se against prostate, lung, and colon carcinogenesis (211). In general, a recent meta-analyses of multiple case–control and randomized studies led to the conclusion that Se is protective against prostate cancer, but only in a relatively narrow range of concentrations (152). Indeed, participants profited from a Se supplementation only when entering the study with a low basal Se status (152). Another meta-analysis of epidemiological data concluded that despite evidence for an inverse association between Se exposure and the risk of some types of cancer, these results should be interpreted with care due to the potential limiting factors of heterogeneity and influences of unknown biases (79). One of these factors is genetic background—a study among smokers demonstrated that depending on a 15-kDa selenoprotein genetic polymorphism, smoking individuals either can benefit from a higher Se intake or have an increased risk for lung cancer (161). Together, the results from a number of ongoing randomized controlled trials (79, 119) should help to better understand the general effect of selenium in cancer.

A. Low-molecular-weight dietary oxidants and antioxidants

The majority of the dietary oxidants are often already intermediates of the prestarted radical reactions. These reactions can be initiated already in the food source and aggravated by storage time or preparation conditions. Virtually, every food, whether it is plant-derived or animal-derived, contains such prooxidants (e.g., ROS themselves or aldehydes). In plants, generation of initial ROS can cause oxidation of linolenate (a prominent component of the plant cell membrane), which is a starting point for producing G₁-phytoprostanes and MDA, while in animals, a similar reaction sequence forms isoprostanes from arachidonate (251). In addition, carbohydrate-rich food processed at high temperature (for instance, French fries or potato chips) are usually containing different amounts of acrylamide, which can act as a prooxidant by depleting GSH (146). Commonly, a number of beverages are also considered as major source of prooxidants; for example, tea and coffee could generate H₂O₂ to achieve over 100 μM concentrations (215).

Although the major human intervention trials failed to confirm the proposed roles of the traditional dietary antioxidants (see section VI), it is worth remembering that there are thousands of plant-derived bioactive compounds that are potentially antioxidants, whereas only half a dozen was tested in those trials. During evolution, plants gained the possibility to generate an enormous number of antioxidant molecules to confront the oxidative challenge they encounter upon photoreception. Thus, plant-derived food is generally considered to be rich in antioxidants. Their amount depends on the plant species, life cycle of the plant, postharvest technologies, and soil type (358). Depending on the method used, the potential antioxidant capacity of different fruits and vegetables varies. However, there is a consensus that cruciferous vegetables, legumes (especially soy origin), sweet potatoes, and tomatoes are some of the most antioxidant-rich vegetables. Berries and many other fruits, for example, apple, citrus, and grapes, hold also strong positions in terms of their antioxidant capacity (98, 291). Due to the tremendous varieties in fruits and vegetables worldwide, the World Cancer Research Fund and the American Institute of Cancer Research recommend serving several different fruits and vegetables instead of only one (115). Although animal-derived food is usually not considered to be rich in antioxidants, their role in the antioxidant system is not negligible. For example, GSH is found in fresh meat (330).

B. Macronutrients and ROS generation/accumulation

The macronutrients stimulate ROS production by increasing the postprandial metabolic rate, especially at the mitochondrial level. Acetyl CoA, which is formed by oxidative conversion of pyruvate (generated from glucose during glucolysis), as well as after β-oxidation of fatty acids and metabolism of several glucogenic amino acids, is processed in mitochondria via the tricarboxylic acid cycle, leading to the generation of NADH and FADH₂ (Fig. 15). NADH and FADH₂ donate electrons to complex I and II, respectively, of the mitochondrial ETC, while leakage of these electrons to oxygen is the main source of O₂ ^•− formation (263). After ingestion of macronutrients, processing of substrates through the ETC is accelerated, and in the absence of sufficient physical activity (i.e., without increased demand for ATP), the mitochondrial membrane potential would be getting more and more reduced, resulting in higher ROS generation (60, 97, 263). Correspondingly, immediately after ingestion of a meal, ROS concentrations in peripheral cells are markedly increased (245, 269). Increased production of ROS would eventually strengthen the existing ROS-mediated activation of various signaling pathways and transcription factors, some of which (i.e., NF-κB, AP-1, and early growth response protein-1 [Egr-1]) promote additional ROS generation (see section IV.B) (10). For instance, ROS-dependent activation of PKC favors NADPH oxidase overexpression (99).

OPEN IN VIEWER

FIG. 15.

Different approaches by which macronutrients modulate ROS generation. Macronutrients (free fatty acids, glucose, and amino acids) stimulate ROS production by increasing the postprandial metabolic rate, especially at the mitochondrial level. In addition, the macronutrients have long-term effects on ROS levels by influencing the activity of transcription factors that regulate the expression of antioxidant and/or ROS-generating enzymes. ETC, electron transport chain; NF-κB, nuclear factor kappa B; NOX, NADPH oxidase subunit; PKC, protein kinase C; PPAR, peroxisome proliferator-activated receptor; SOD, superoxide dismutase; TCA, tricarboxylic acid.

A. Type, dose, combination, and duration of exposure to antioxidants determine their effects

The reasons for the lack of protective effects of antioxidants in human trials are not known. Among the possible explanations are the wrong type, dose, combination, and/or duration of exposure. Indeed, in almost all of the large-scale trials with antioxidants, a significantly high dose of each antioxidant was applied either alone or in limited combinations. The question in these studies was only whether a specific amount and combination of antioxidants can prevent tumorigenesis in a specific randomized population, but mechanisms of action have drawn less attention. In this respect, it should be remembered that some antioxidants, including vitamin E, vitamin C, and quercetin, can have prooxidative properties at high concentrations. Meta-analysis of the dose–response relationship between vitamin E supplementation and total mortality indicates that administration of α-tocopherol at high doses (≥400 IU/day) may be associated with increased mortality rate, an observation attributed to the prooxidant effects of α-tocopherol at these concentrations (239). It was also shown that β-carotene acts as a prooxidant in the presence of chronic oxidative stress such as that resulting from smoking (149, 229). In line, quercetin at concentrations in the range of 1–40 μM reacts with ROS and chelates metal ions, whereas at concentrations higher than 40 μM, it increases oxidative stress (362).

Moreover, one antioxidant component can stipulate another's presence or even block its action. For example, supplementation of mildly hypercholesterolemic subjects with vitamin E lowered the plasma levels of the antioxidant ubiquinol-10 (172). On the other hand, the beneficial effects of multiple dietary antioxidants combined may be greater than the effect of each of those compounds examined individually (119). In particular, the natural combination of phytochemicals and vitamins in fruits and vegetables is believed to be responsible for their potent antioxidant activity (213). For example, it was calculated that vitamin C in apples contributes only to about 0.4% of their TAC (92).

Since neoplastic lesions are multistage processes and may take several decades to become clinically detectable, the duration of antioxidant therapy in human trials, though always at least several years, could have been insufficient for evaluating their efficiency in preventing cancer. Indeed, in colorectal cancer prevention study performed by the American Cancer Society that lasted for 5 years, it was found that a regular multivitamin use at enrollment was not associated with the risk of colorectal cancer, whereas regular multivitamin use 10 years before enrollment was associated with a reduced risk, no matter whether participants still used multivitamins at enrollment or had stopped (162). These results are consistent with the hypothesis that a past, but not recent, vitamin use may be associated with a reduced risk of colorectal cancer, though it should be taken into account that multivitamin users were more likely than nonusers to practice health conscious behaviors in general, including exercising more (162). The effects of a recent antioxidant use may only be detectable at a subclinical level, whereas a reduction in the risk of clinically detectable disease may require longer exposures. Therefore, a useful alternative to the clinical endpoints is the use of carcinogenesis biomarkers, cell cycle markers, and markers of oxidative stress/inflammation (119). In line, one recent study showed that an antioxidant micronutrient cocktail (800 mg dl-α-tocopherol acetate, 24 mg β-carotene, 1.0 g vitamin C, 200 μg L-selenomethionine, 7.2 mg riboflavin, 80 mg niacin, 60 mg zinc, and 5 mg manganese) given daily over 4 months can modulate the biomarkers of oxidative stress and inflammation in humans (149). However, only development of better methods for quantitative detection of various ROS in vivo (see section III.C) may answer the question whether antioxidants in clinical trials failed to reduce/abolish oxidative damage, or they indeed reduced oxidative damage without any beneficial effects, if the latter is correct that would imply that oxidative damage only correlates with cancer, but is not causative for it (71).

B. The sensitivity to the antioxidant action has individual variations

Not only the dose, duration, and/or the way of antioxidant administration, but also individual nutritional as well as the health status of participants can strongly influence the outcome of trials. The Linxian study that corroborated to the protective role of a combination of antioxidants was carried out with a severely undernourished population. On the other hand, many enrolled subjects in the Western countries consumed antioxidants independently of participation in trials. The lack of effects of antioxidants in the Italian population may be explained by the simple suggestion that Italians already receive the benefits of a Mediterranean diet rich in phenolic antioxidants; thus, this population may be selected not to respond in a beneficial way to antioxidant supplements (131). In general, there are no large clinical trials in which only patients that demonstrate elevated ROS levels were recruited, though only these subjects are expected to benefit from the action of antioxidants. In some particular subgroups of patients, the positive effect of antioxidants was indeed shown. For example, the combination of γ-tocopherol and vitamin C significantly reduced the progression of atherosclerosis in hypercholesterolemic smoking men who had low baseline serum vitamin E levels (310).

Single-nucleotide polymorphisms in genes encoding important enzymes regulating the oxidative status of a cell can also contribute to enhanced oxidative stress in some individuals. For instance, a genetic polymorphism in the MnSOD gene can potentially cause impaired protein targeting and is found to be associated with prostate (203), breast (12), and ovarian (270) cancer. Similarly, differences in GST P activity caused by a genetic polymorphism have been found to correlate with oxidative DNA damage and breast cancer (227). Interestingly, various studies have revealed that genetic variations in several of the antioxidant-related genes are very common throughout the world and may vary among different ethnic and racial communities (1, 31, 135, 188). Since antioxidant supplementation trials did not take into account the single-nucleotide polymorphisms within the genes for the antioxidant enzymes in the corresponding populations, this could be one of the explanations of these trials' contradictory results.

C. The effects upon ROS removal differ depending on their type, concentration, and intracellular localization

The failure of antioxidant trials can be also viewed as evidence against reductionism in science, especially in such a multifaceted connection between ROS and cancer. The research of the last decade indicated that there is a tremendously wide array of mechanisms through which ROS could contribute to tumorigenesis. While direct damage of DNA molecules by ROS and induction of epigenetic changes can contribute to a genetic program that is favorable for tumorigenesis, it became also clear that the effect of ROS often depends on the stage of tumorigenesis. In this regard, it should be noticed that several studies involved smokers, a portion of whom could have suffered from the subclinical stage of cancer. Since many antitumor agents act by inducing ROS production in cancer cells, supplementation of smokers with high doses of antioxidants could be harmful, and it may partially explain the adverse effects observed in ATBC or CARET. Further, the protective or tumorigenic effects could also depend on the type of ROS involved. For example, O₂ ^−• and H₂O₂ downregulated caveolin-1 expression and inhibited cell migration and invasion, whereas •OH upregulated caveolin-1 expression and promoted cell migration and invasion of the human lung carcinoma H460 cells (218).

Increasing evidence suggests that a number of antioxidant compounds, which were previously thought to have antitumor properties due to their ROS scavenging activity, are more likely to work through cell-to-cell signaling, inflammatory enzyme activity, or gene regulation (252, 261, 299, 368). New data also indicate that some substances that were thought to be carcinogenic due to the stimulation of oxidative stress may indeed act via other mechanisms (13). For example, the prolonged administration of peroxisome proliferators, which include some hypolipidemic drugs (clofibrate), herbicides, plasticizers, and leukotriene antagonists, leads to development of hepatic carcinomas in rodents. Peroxisome proliferators sustain a PPARα-dependent activation of the H₂O₂-generating enzymes, aldehyde oxidase-1 and cytochrome P450 4A, and a decline in the activity of H₂O₂-degrading catalase and GPX, leading to the accumulation of the oxidative stress marker lipofuscin and the formation of 8-oxodG–DNA adducts in liver cells (178, 292). However, the current view favors the idea that hepatic carcinogenesis promoted by peroxisome proliferators is not only due to hepatic oxidative stress but also related to the abnormalities in cell growth and communication (13, 59).

It is also possible that orally administered antioxidants may be inaccessible to the source of free radicals in various cellular compartments. For most cell types, antioxidant molecules to be effective should be specifically localized either in the plasma membrane, cytoplasm, mitochondria, peroxisomes, nuclei, or other organelles (see sections II and V.A). For example, water-soluble molecules, such as vitamin C, are radical scavenging agents in the cytoplasm, whereas lipid-soluble molecules, such as vitamin E, β-carotene, and polyphenols, appear to act primarily as antioxidants within the lipid environments. The slow rate of tocopherol movement between the bilayer leaflets could imply that in the absence of an active uptake mechanism, exogenously applied tocopherols stay in the outer leaflet of the cell membranes (18). Therefore, lipid-soluble antioxidants have a low ability to scavenge ROS produced in the cytoplasm and various intracellular organelles (144). Although vitamin C has better access to the intracellular compartments as compared to lipid-soluble vitamins, it has a narrow absorption window with a maximum at 200 mg, and therefore an increase in the orally administered dose will not increase its bioavailability (235).

Resveratrol can behave either as an antioxidant or as a prooxidant depending on the selective intracellular microenvironment. As an antioxidant, resveratrol, in particular, decreases mitochondrial production of O₂ ^•− in various cell types (355). It was recently shown that the prooxidant activity of resveratrol is dependent on its interaction with chromatin-bound copper that leads to ROS generation through Fenton-type reactions with subsequent oxidative DNA breakage and apoptosis (252). Such resveratrol-induced DNA damage is enhanced in low pH environments characteristically observed in tumors due to the Warburg effect. Wrong intracellular localization of antioxidants can be also harmful (131). For example, correct disulfide-bridge formation and folding of proteins in the ER are aided by having an environment more oxidized than the rest of the cell; therefore, an increase of antioxidants in the ER may disrupt the localized oxidation necessary for its physiological function (121). Thus, when using antioxidants, their different action in various cellular compartments as well as their influence on different signaling cascades needs to be considered.

D. Diet, oxidative stress, and cancer

Still, the protective effects of antioxidant-rich food with respect to tumorigenesis are widely reported (2, 183, 394). A selected dietary fashion can result in less ROS generation and provide us with less vulnerability to oxidative stress. These beneficial effects could be achieved primarily by dietary intervention, but not by the micronutrient supplementation (213). Dietary intake does not involve pharmacologic doses of a single constituent and is unlikely to elicit adverse effects such as a high-dose supplementation of various nutrients does. Most importantly, the form of a specific antioxidant present in normal diet could be a vital factor that specifically contributes to its function, and might be overlooked in supplementation studies where the source and form vary from study to study. In line, while pharmacological doses of single antioxidants have demonstrated harmful effects in smokers, two food-based antioxidant-rich diets were found to be safe in a randomized controlled trial in male smokers (176). As a new tool for investigating the relationship between diet and cancer risk in population studies, the measurement of plasma TAC has been proposed (325). A comprehensive food database consisting of the TAC for typical foods as well as other dietary items such as traditional medicine plants, herbs, and spices (more than 3100 in general) was recently published and can be used in future investigations on diet and cancer (55). Recent data of the European Prospective Investigation into Cancer (EPIC) study, including a total of 521,457 subjects, indicate that a high intake of antioxidant compounds from different sources of plant foods as measured by TAC, assessed by TRAP and FRAP assays, is associated with a reduction in the risk of gastric cancer (326). This promising result highlights the importance of synergistic interactions between plant food components as compared to the use of single antioxidants.

As an example, we would shortly discuss the diets of two healthiest communities of the world: the Okinawans and the Mediterranean basin dwellers (39, 378). The Okinawan diet is in principle a low caloric one with the total daily caloric intake as low as 1100 kcal. In a typical Okinawan diet, the carbohydrate intake accounts for 85% of total calorie, while protein and fat intake are 9% and 6%, respectively (378). Another feature of traditional Okinawan diet is the high amount of fruits and vegetables in their cuisine, which are highly rich in antioxidants, for example, sweet potato, legumes, soy products, bitter melon, and seaweeds (378). Fish holds a moderate position in the traditional Okinawan diet, and meat consumption is significantly lower than in the rest of Japan. Clearly, the traditional Okinawan diet promises lesser threat of oxidative stress and oxidative stress-induced tumorigenesis. However, dietary changes were observed in the traditional Okinawan diet for some decades associated with a marked increase in the mortality rate due to cancer (243, 378).

The Cretan diet represents the traditional Mediterranean diet of the 1960s, and the population consumes higher amount of fat (37% of total energy) than the Okinawan population, which is actually striking, given their second lowest rate of incidence of cancer or cardiovascular disease just behind the Okinawans (330). However, it seems to be likely the fatty acid composition holds the key. A large portion of fat in the Cretan diet comes from olive oil, fish, and chicken. Olive oil is rich in monounsaturated fatty acid (MUFA) instead of PUFA; fish oil is one of the higher sources of ω-3 fatty acids, and the chicken contains more ω-3 fatty acids because of being grass- and worm-fed in Crete (330). As a result, the fatty acids are mainly MUFA, and due to this, the ω-3/ω-6 fatty acid ratio is extremely high (329). Another feature of the Cretan diet is the regular consumption of a high amount of fruits considered to be rich in antioxidants. Unlike in the Okinawan population, wine consumption is a common dietary pattern in Crete, which is thought to provide them with resveratrol. Although the beneficial effect of resveratrol is not yet well proven, it is extensively reported to be a potential antioxidant and antitumorigenic phytochemical (see section VI.D) (370).

However, it is also not always clear to what extent the dietary fashion could provide antioxidant support. Antioxidant molecules' absorption, bioavailability, and function are found to be highly dependent on the food microstructure. Several food components can be responsible for limited bioaccessibility and bioavailability of antioxidants (275). It is also imperative to unravel whether any food component either directly or indirectly promotes antioxidant absorption that might present a stumbling block to antioxidant supplementation.