Interactions Between Antioxidants: How Useful Are In Vitro Studies for Understanding the Action of Antioxidants in Complex Systems?

Abstract

Interactions between antioxidants are believed to be fundamental for proper nutrition and health benefits. This review discusses various types of antioxidant interactions in vitro, their mechanisms, and their dependence on assay conditions and evaluates the validity of extrapolations of results obtained from cell-free assays and assays of cellular antioxidant activity to in vivo and food systems. In vitro studies revealed that the type of interaction (additive, antagonistic, or synergistic) depends not only on the identity of compounds but also on their absolute concentrations, concentration ratio, type of assay, reaction medium, and the method of analysis of results. Moreover, reactions of antioxidants in simple model systems, employing synthetic indicators, do not fully reflect their reactivity in food systems and organisms. Although in vitro studies may be a useful and necessary step for explaining the mechanisms of interactions between antioxidants, their importance for explaining the phenomena observed in vivo seems limited, as the organismal effects of compounds referred to as antioxidants depend rather on their actions on specific signaling pathways than on their direct antioxidant action. Nevertheless, they may provide useful information for food preservation, allowing for the reduction of the amounts of antioxidant additives.

1. Introduction

In recent decades, antioxidants have aroused broad interest in both biomedicine and food science. This interest was partly due to the popularization of the free radical theory of aging, postulating that antioxidants can slow down the aging process [1,2,3], to the beneficial effects of the antioxidant-rich Mediterranean diet [4,5,6], to the occurrence of oxidative stress in many diseases [7,8,9]. In food science, the increasing use of processed food has augmented the demand for safe natural antioxidants [10,11,12].

Studies of individual antioxidants have made it possible to understand the mechanisms of their action at various levels. However, both in organisms and food products, antioxidants are not present individually but coexist with a plethora of other antioxidants, metabolites, proteins, lipids, minerals, and other matrix components. Therefore, interactions among antioxidants are of interest, primarily to understand their action in complex systems.

The null hypothesis concerning the behavior of several antioxidants in a system is that their combined effect is equal to the sum of individual effects (additivity). Such behavior is indeed observed in many systems. However, in other cases, the effect of a mixture of antioxidants is stronger than the sum of the effects of individual antioxidants (synergy) or weaker than the sum of individual components (antagonism). This classification seems intuitively obvious; however, in practice, diagnosis of the type of interaction depends on the mathematical model used to describe the interaction, and distinguishing between additive and non-additive interactions is often more or less arbitrary.

Usually, consuming whole fruits or vegetables shows more beneficial effects than consuming individual dietary supplements, which is widely explained by the interactive effects of coexisting phytochemicals in whole foods [13,14,15]. Interactions between antioxidants may have a practical dimension in food and other systems prone to oxidation, as synergy between antioxidants may reduce the amount of antioxidant additions necessary to prevent oxidation [16,17,18].

An insight into the in vivo interactions between antioxidants is difficult. Therefore, interactions between antioxidants are mainly studied in model in vitro systems, enabling studies of pairs or groups of antioxidants. Can results of such studies be extrapolated to in vivo systems or complex food products? Evaluation of the validity of such extrapolations of results obtained from cell-free assays and assays of cellular antioxidant activity is the aim of the present review.

2. Assessment of the Type of Interaction

Antioxidant reactions include electron transfer, hydrogen atom transfer, radical addition, or complexation [19] and can interact with each other. The type of interaction between antioxidants is assessed by comparison of a parameter of an antioxidant assay (e.g., extent of reduction of an indicator, lag time of the reaction) of individual antioxidants and of their mixture. Often, if a linear relationship between the antioxidant activity and the antioxidant concentration is expected, the results of assays are directly compared (direct summation), which is sometimes referred to as the “traditionally calculated mixture effect” [20]. In this approach, if the effect of a mixture of two or more antioxidants coincides (within the limits of experimental error) with the sum of separate effects of components of the mixture, additivity is diagnosed; if the effect of the mixture is higher/lower than the sum of effects of individual components, their interaction is classified as synergy or antagonism, respectively. An important limitation of this approach is the necessity of a strict proportionality between the amounts of antioxidants and the observed effect, which is generally not fulfilled in the ABTS^• and DPPH^• decolorization, β-carotene bleaching, and ORAC assays. Other models describing interactions imply other criteria of additivity, antagonism, and synergy.

The Chou–Talalay method is one of the most widely used methods for detecting and quantifying synergistic interactions between two or more drugs (for which it was originally developed) or other compounds, including antioxidants. In this approach, the main equation forming the basis of this method was derived as a unified theory of the Michaelis–Menten, Hill, Henderson–Hasselbalch, and Scatchard equations, known as the median-effects equation:

$${\displaystyle \frac{f_a}{f_u}}= {\left({{\displaystyle \frac{D}{D}}}_m\right)}^m$$ (1)

where f_a is the effect (e.g., fraction of an indicator reduced by an antioxidant), f_u is the fraction of indicator unaffected, D is the dose of an acting antioxidant, D_m is the median-effect dose, and m is the sigmoidicity of the dose–effect curve. This equation can be linearized:

$$log\left({\displaystyle \frac{f_a}{f_u}}\right)= m \mathit{log}\left(\mathrm{D}\right) - m \mathit{log} \left(D_m\right)$$ (2)

From this form of the equation, the values for D_m and _m can be estimated. These values can then be used to calculate the effect of the interaction between compounds (antioxidants) using the Combination Index CI:

$$\mathrm{C}\mathrm{I}= {\displaystyle \frac{D_1}{E_1}}+ {\displaystyle \frac{D_2}{E_2}}$$ (3)

where CI is the effect (e. g. extent of reduction of an indicator) observed for the binary mixture, D₁ and D₂ represent the effect at the single doses of compound A and compound B, and E₁ and E₂ are theoretical doses of the compounds expected to achieve the experimentally measured response. E₁ and E₂ can be calculated using the previously computed D_m and m values. CI < 1 indicates synergy, CI = 1 indicates additivity, and CI > 1 indicates antagonism [20,21,22,23,24].

This method also has its limitations. With adequate experimental accuracy of measurements, a minimum of only 10 data points is required for quantitative synergy determination in two-compound combinations [24]. Non-linearity of the dose–response curves, which occurs frequently, may lead to erroneous conclusions on the types of interactions [25].

When the concentration dependence of the reaction of an antioxidant with an indicator is hyperbolic, the method of Webb [26] for the estimation of additivity of effects, derived for enzyme inhibition, can be used:

Expected reduction (fraction) = 1 − [1 − Reduction by Antioxidant 1 (fraction)] × [1 − Reduction by Antioxidant 2 (fraction)] (4)

This approach was used, i.a, to estimate interactions between antioxidants in the DPPH^• decolorization assay [27] and in the ABTS^• decolorization assay [20].

An obvious limitation of all these methods consists in the exactness of matching experimental data with the formula used. If, for any reason, deviation from the expected dependence occurs, the prediction of the combined effect will be charged with error.

A general problem for all methods of evaluation of interactions is the difficulty in distinguishing between additive and non-additive effects when the measured antioxidant activity only slightly deviates from the 100% expected for additivity, or the CI slightly differs from 1. In the “traditionally calculated mixture effect”, a criterion of 10% or even 5% difference [28] from 100% additivity is sometimes arbitrarily used, but usually the statistical significance of differences from 100% additivity is evaluated. As CI is not associated with a standardized statistic or p value, designations of significant synergy or significant antagonism must be determined by largely arbitrary thresholds. Chou specifies a narrow range of CI values (0.9–1.1) as “nearly additive” but offers no physical, statistical, or theoretical rationale for those thresholds [29]. More sophisticated statistical methods for determination of confidence intervals have also been proposed [30].

Two examples may illustrate the importance of criteria of differentiation of the type of interaction between antioxidants: the “10% criterion” can lead to different conclusions than statistical evaluation [20,26], and various methods of analysis of interaction can yield divergent conclusions on the types of interactions [20].

Although the statistical evaluation of differences from additivity should be recommended, it is difficult to avoid the impression that the precision of measurements, rather than the real difference in the nature of the interaction, is the discriminating factor in the classification. For this reason, small though statistically significant deviations from additivity are not necessarily an indication of a chemically or biologically meaningful synergy.

3. The Type of Interaction Depends on Many Variables

The most common assays used in studies of interactions between antioxidants and their mechanisms, reaction conditions, limitations, and physiological relevance are listed in Table 1.

Table 1.

Mechanisms, conditions, limitations, and physiological relevance of main assays used in studies of antioxidant interactions.

Assay (Mechanism)	Oxidant/Indicator	Medium	pH	Major Limitations	Physiological Relevance	References
ABTS• decolorization (SPLET)	ABTS•	Aqueous or ethanol	Usually 7.4 but may be lower	Limited reactivity of hydrophobic antioxidants; short reaction times underestimate the activity of slow-reacting antioxidants	Limited (non-physiological substrate)	[31,32,33]
DPPH• decolorization (SPLET)	DPPH•	Methanol or ethanol	Usually not controlled	Limited solubility of some antioxidants; insolubility of proteins. Colored compound may interfere	Limited (non-physiological substrate)	[34,35,36]
FRAP (SET)	Fe(III)	Acetate buffer	3.6	Low reactivity of thiols. Colored compound may interfere	Limited (unphysiological pH, Fe(III) is not a major physiological oxidant)	[37,38]
ORAC (HAT)	Usually APPH/fluorescein	Phosphate buffer	Usually 7.0–7.4	Needs a fluorimeter or fluorimetry reader; time-consuming	Physiologically relevant (measures reactivity for peroxyl radicals, measurement usually at 37 °C)	[39,40,41]
β-Carotene bleaching (HAT)	Spontaneous or AAPH, metal ions, or heme/Carotene	Aqueous buffer + detergent	Variable, usually 7.0–7.4	Various protocols, time-consuming	Physiologically relevant (measures reactivity for peroxyl radicals)	[42,43,44]
Inhibition of oxidation of linoleic acid, oils or other fats (HAT, RAF)	Spontaneous or AAPH, metal ions, or heme/O2 uptake, conjugated dienes or Rancimat	Aqueous buffer + detergent	Variable, usually 7.0–7.4	Various protocols, time-consuming	Physiologically relevant (measures reactivity for peroxyl radicals)	[45,46,47]
Cellular antioxidant activity	Usually AAPH/fluorogenic ROS-sensitive probes	Cells pre-loaded with antioxidants	Usually 7.4	Interference from endogenous antioxidants, intracellular concentration of exogenous antioxidants unknown; cell-type dependence	Physiologically relevant (measures antioxidant activity in cells)	[48,49,50]

HAT, hydrogen electron transfer; RAF, radical adduct formation; SET, single electron transfer; SPLET, sequential proton loss electron transfer.

An extensive review of synergistic, antagonistic, and additive antioxidant interactions in binary antioxidant mixtures in vitro has been published [51], so only exemplary results are reported here (Table 2 and Table 3).

Table 2.

Exemplary quantitative results of interactions between antioxidants in vitro.

Antioxidant System	Assay	Interaction Parameter = 100% × [(Effect Measured)/(Effect Predicted as the Arithmetic Sum of Individual Antioxidant Activities)]; Type of Interaction		Reference
3.8 µM Taxifolin + 4.2 µM GSH	ABTS• decolorization, “classical” analysis	88.9	Ant T	[20]
1.8 µM Taxifolin + 9.4 µM GSH	ABTS• decolorization, “classical” analysis	90.1	Add T	[20]
1.2 µM Taxifolin + 11.9 µM GSH	ABTS• decolorization, “classical” analysis	91.3	Ant T	[20]
0.7 µM Taxifolin + 11.1 µM GSH	ABTS• decolorization, “classical” analysis	91.5	Ant T	[20]
3.8 µM Taxifolin + 4.2 µM GSH	ABTS• decolorization, Webb’s simulation	102.9	Add T	[20]
1.8 µM Taxifolin + 9.4 µM GSH	ABTS• decolorization, Webb’s simulation	102.9	Add T	[20]
1.2 µM Taxifolin + 11.9 µM GSH	ABTS• decolorization, Webb’s simulation	101.5	Add T	[20]
0.7 µM Taxifolin + 11.1 µM GSH	ABTS• decolorization, Webb’s simulation	97.8	Add T	[20]
2.3 µM Quercetin + 2.6 µM GSH	ABTS• decolorization, “classical” analysis	97.1	Add T	[20]
1.3 µM Quercetin + 6.5 µM GSH	ABTS• decolorization, “classical” analysis	92.9	Ant T	[20]
0.8 µM Quercetin + 7.7 µM GSH	ABTS• decolorization, “classical” analysis	94.4	Add T	[20]
0.7 µM Quercetin + 10 µM GSH	ABTS• decolorization, “classical” analysis	89.5	Add T	[20]
2.3 µM Quercetin + 2.6 µM GSH	ABTS• decolorization, Webb’s simulation	110.6	Syn T	[20]
1.3 µM Quercetin + 6.5 µM GSH	ABTS• decolorization, Webb’s simulation	104.8	Add T	[20]
0.8 µM Quercetin + 7.7 µM GSH	ABTS• decolorization, Webb’s simulation	103.1	Add T	[20]
0.7 µM Quercetin + 10 µM GSH	ABTS• decolorization, Webb’s simulation	97.7	Add T	[20]
0.1 mM Gallic acid + 0.1 mM procatechuic acid	DPPH• decolorization	141.5	Syn A	[21]
0.1 mM Gallic acid + 0.1 mM chlorogenic acid	DPPH• decolorization	102.1 #	Syn A	[52]
0.1 mM Protocatechuic acid + 0.1 mM chlorogenic acid	DPPH• decolorization	112.9 #	Syn A	[52]
0.1 mM Gallic acid + 0.1 mM vanillic acid	DPPH• decolorization	91.7 #	Ant A	[52]
0.1 mM Protocatechuic acid + 0.1 mM vanillic acid	DPPH• decolorization	103.5 #	Add A	[52]
66 µM Gallic acid + 66 µM chlorogenic acid + 66 µM vanillic acid	DPPH• decolorization	177.6 #A	Syn A	[52]
66 µM Gallic acid + 66 µM procatechuic acid + 66 µM chlorogenic acid	DPPH• decolorization	106.3 #A	Syn A	[52]
66 µM Gallic acid + 66 µM procatechuic acid + 66 µM vanillic acid	DPPH• decolorization	120.8 #A	Syn A	[52]
66 µM Protocatechuic acid + 66 µM chlorogenic acid + 66 µM vanillic acid	DPPH• decolorization	82.0 #A	Ant A	[52]
Caffeic acid + α-tocopherol:	DPPH• decolorization			[53]
0.2:1		144	Syn N10
0.5:1		159	Syn N10
1:1		133	Syn N10
2:1		138	Syn N10
Quercetin + α-tocopherol:	DPPH• decolorization			[53]
0.5:1		180	Syn N10
1:1		118	Syn N10
2:1		116	Syn N10
5:1		163	Syn N10
2.94 µM Gallic acid + 11.8 µM caffeic acid	FRAP	237.8	Syn N10	[16]
2.94 µM Rosmarinic acid + 11.8 µM caffeic acid	FRAP	137.5	Syn N10	[16]
2.94 µM Quercetin + 2.94 µM rutin	FRAP	106.3	Add N10	[16]
2.94 µM Quercetin + 2.94 µM rutin + 11.8 µM caffeic acid	FRAP	159.4	Syn N10	[16]
2.94 µM Quercetin + 2.94 µM rutin + 2.94 µM rosmarinic acid	FRAP	155.2	Syn N10	[16]
3.2 µM p-Coumaric acid + 3.2 µM sinapic acid	FRAP	96.0	Add N5	[28]
16.1 µM p-Coumaric acid + 16.1 µM sinapic acid	FRAP	108.0 #	Syn N5	[28]
32.3 µM p-Coumaric acid + 32.3 µM sinapic acid	FRAP	118.0 #	Syn N5	[28]
3.2 µM Caffeic acid + 3.2 µM rosmarinic acid	FRAP	96.2 #	Add N5	[28]
16.1 µM Caffeic acid + 16.1 µM rosmarinic acid	FRAP	112.1 #	Syn N5	[28]
32.3 µM Caffeic acid + 32.3 µM rosmarinic acid	FRAP	115.8 #	Syn N5	[28]
3.2 µM p-Coumaric acid + 3.2 µM caffeic acid + 3.2 µM rosmarinic acid	FRAP	226.6 #	Syn N5	[28]
16.1 µM p-Coumaric acid + 16.1 µM caffeic acid + 16.1 µM rosmarinic acid	FRAP	147.6 #	Syn N5	[28]
32.3 µM p-Coumaric acid + 32.3 µM caffeic acid + 32.3 µM rosmarinic acid	FRAP	110.8 #	Syn N5	[28]
0.625 µM p-Coumaric acid + 0.625 µM ferulic acid	ORAC	410.3 #	Syn N5	[28]
0.625 µM Caffeic acid + 0.625 µM sinapic acid	ORAC	311.5 #	Syn N5	[28]
0.625 µM Gentisic acid + 0.625 µM syringic acid	ORAC	336.2 #	Syn N5	[28]
0.625 µM Vanillic acid + 0.625 µM syringic acid	ORAC	310.4 #	Syn N5	[28]
0.625 µM Protocatechuic acid + 0.625 µM gentisic acid + 0.625 µM vanillic acid	ORAC	149.1 #	Syn N5	[28]
0.625 µM Protocatechuic acid + 0.625 µM gentisic acid + 0.625 µM vanillic acid + 0.625 µM syringic acid	ORAC	165.4 #	Syn N5	[28]
0.625 µM Ferulic acid + 0.625 µM sinapic acid + 0.625 µM rosmarinic acid	ORAC	77.0 #	Ant N5	[28]
0.625 µM Rosmarinic acid + 0.625 µM p-coumaric acid + 0625 µM ferulic acid + 0.625 µM sinapic acid	ORAC	142.6 #	Syn N5	[28]
0.625 µM Protocatechuic acid + 0.625 µM gentisic acid + 0.625 µM gallic acid 0.625 µM vanillic acid + 0.625 µM syringic acid	ORAC	127.7 #	Syn N5	[28]
0.625 µM p-Coumaric acid + 0.625 µM caffeic acid + 0.625 µM ferulic acid 0.625 µM sinapic acid + 0.625 µM rosmarinic acid	ORAC	101.8 #	Add N5	[28]
Quercetin + resveratrol, 100 µg/mL	TRAP	64.9 #	Ant N10	[54]
Caffeic acid + α-tocopherol:	β-Carotene bleaching			[53]
0.2:1		96	Add N10
0.5:1		96	Add N10
1:1		97	Add N10
2:1		95	Add N10
Quercetin + α-tocopherol:	β-Carotene bleaching			[53]
0.5:1		163	Syn N10
1:1		109	Add N10
2:1		91	Add N10
5:1		108	Add N10
Up to 5 µM α-Tocopherol+ equimolar caffeic acid	Inhibition of AAPH-induced oxidation of linoleic acid, lag time	81	Ant N10	[55]
Up to 11 µM Quercetin + equimolar rosmarinic acid	Inhibition of AAPH-induced oxidation of linoleic acid, lag time	113	Syn N10	[55]
4.5 µM Astaxanthin + 4.5 µM β-carotene	Inhibition of AMVN-induced oxidation of linoleic acid, lag time	146.8 #	Syn N10	[56]
4.5 µM Astaxanthin + 4.5 µM lycopene	Inhibition of AMVN-induced oxidation of linoleic acid, lag time	160.5 #	SynN10	[56]
4.5 µM Astaxanthin + 4.5 µM zeaxanthin	Inhibition of AMVN-induced oxidation of linoleic acid, lag time	93.8 #	Add N10	[56]
1 mM Curcumin + 1 mM (−)epicatechin	Inhibition of sunflower oil triacylglycerol autoxidation at 80 °C	102.4 #	Add T	[57]
1 µM Resveratrol + 1 µM PMHC, pH 4.0	Inhibition of AAPH-induced DMPC oxidation, lag time	114.1 #	Syn N10	[58]
1 µM Resveratrol + 1 µM PMHC, pH 6.0	Inhibition of AAPH-induced oxidation of methyl linoleate in DMPC liposomes, lag time	116.2 #	Syn N10	[58]
1 µM Resveratrol + 1 µM PMHC, pH 7.0	Inhibition of AAPH-induced oxidation of methyl linoleate in DMPC liposomes, lag time	174.4 #	Syn N10	[58]
1 µM Resveratrol + 1 µM PMHC, pH 8.0	Inhibition of AAPH-induced oxidation of methyl linoleate in DMPC liposomes, lag time	91.0 #	Add N10	[58]
7.5 µM α-Tocopherol + 10 µM (−)epigallocatechin gallate + 10 µM ascorbic acid	Oxidation of linoleic acid in SDS micelles, lag time	200.0 #	Syn N10	[59]
7.5 µM α-Tocopherol + 10 µM (−)epicatechin gallate + 10 µM ascorbic acid	Oxidation of linoleic acid in SDS micelles, lag time	143.5 #	Syn N10	[59]
7.5 µM α-Tocopherol + 10 µM (−)epigallocatechin + 10 µM ascorbic acid	Oxidation of linoleic acid in SDS micelles, lag time	147.1 #	Syn N10	[59]
77 µM Ascorbic acid + 77 µM chlorogenic acid + 77 µM cysteine	Superoxide scavenging EPR, DMPO spin trap	104.5 #	Add N10	[60]
50 µM Ascorbic acid + 50 µM chlorogenic acid + 50 µM cysteine	t-Butyl peroxide scavenging EPR, DMPO spin trap	265.0 #	Syn N10	[60]
476 µM Ascorbic acid + 476 µM chlorogenic acid	Photoelectrochemical sensor, amperometric	235	Syn N10	[61]
476 µM Ascorbic acid + 476 µM gallic acid	Photoelectrochemical sensor, amperometric	195	Syn N10	[61]
476 µM Ascorbic acid + 476 µM proanthocyanidins	Photoelectrochemical sensor, amperometric	186	Syn N10	[61]
8 µM * (Lycopene + quercetin), Molar ratio 1:10	Cellular antioxidant activity, HUVECs	120.2 #	Syn A	[62]
8 µM * (Lycopene + quercetin), Molar ratio 1:5	Cellular antioxidant activity, HUVECs	147.9 #	Syn A	[62]
8 µM * (Lycopene + quercetin), Molar ratio 1:1	Cellular antioxidant activity, HUVECs	160.9 #	Syn A	[62]
8 µM * (Lycopene + quercetin), Molar ratio 5:1	Cellular antioxidant activity, HUVECs	201.9 #	Syn A	[62]
8 µM * (Lycopene + quercetin), Molar ratio 10:1	Cellular antioxidant activity, HUVECs	164.7 #	Syn A	[62]
8 µM * (Lycopene + quercetin), Molar ratio 1:10	Cellular antioxidant activity, L-02 cells	112.4 #	Syn A	[62]
8 µM * (Lycopene + quercetin), Molar ratio 1:5	Cellular antioxidant activity, L-02 cells	117.7 #	Syn A	[62]
8 µM * (Lycopene + quercetin), Molar ratio 1:1	Cellular antioxidant activity, L-02 cells	100.7 #	Add A	[62]
8 µM * (Lycopene + quercetin), Molar ratio 5:1	Cellular antioxidant activity, L-02 cells	79.7 #	Ant A	[62]
8 µM * (Lycopene + quercetin), Molar ratio 10:1	Cellular antioxidant activity, L-02 cells	93.2 #	Add A	[62]

The concentrations indicated are final concentrations in the reaction mixture. ^# Interaction parameter calculated based on primary data presented in a publication. AAPH, 2,2′-azobis(2-amidinopropane) dihydrochloride; AMVN, 2,2′-azobis(2,4-dimethylvaleronitrile); DMPC, dimirystoylphosphatidylcholine; DMPO, 5,5-dimethyl-1-pyrroline N-oxide; GSH, glutathione; HUVECs, human umbilical vein endothelial cells; PMHC, 2,2,5,7,8-ppentamethyl-6-hydrochromanol. Add, additivity; Ant, antagonism; Syn, Synergy. ^A, statistical evaluation, one-way ANOVA; ^T; statistical evaluation, Student’s “t” test; ^N10, no statistical evaluation; the criterion of 10% assumed; ^N5, no statistical evaluation; the criterion of 5% assumed. * Concentration used for cell treatment; intracellular concentration not determined.

Table 3.

Exemplary qualitative results of interactions between antioxidants in vitro.

System	Assay	Effects	Reference
Taxifolin + GSH, quercetin + GSH, rutin + GSH, morin + GSH	ABTS• decolorization	Antagonism or additivity (“classic” analysis) Synergy or additivity Webb’s simulation) T	[20]
Chlorogenic acid + gallic acid	ABTS• decolorization	Synergy (111–135% of additive activity) A	[63]
Caffeic acid + gallic acid	ABTS• decolorization	Synergy (113–142% of additive activity) A	[63]
S-Allyl-L-cysteine + catechin	ABTS• decolorization	Antagonism (CI > 1) D	[64]
S-Allyl-L-cysteine + quercetin		Antagonism (CI > 1) D
S-Allyl-L-cysteine + caffeic acid		Antagonism (CI > 1) D
S-Allyl-L-cysteine + ferulic acid		Antagonism (CI > 1) D
S-Allyl-L-cysteine + sinapic acid		Antagonism (CI > 1) D
S-Allyl-L-cysteine + 3,4-dihydroxybenzoic acid		Antagonism (CI > 1) D
Curcumin + quercetin	ABTS• decolorization, DPPH• decolorization	Antagonism, decreasing with increasing the quercetin/curcumin ratio from 1:4 to 4:1 A	[65]
Rutin + resveratrol 2:1, 3.3–16.7 µM	DPPH• decolorization	Antagonistic, % decolorization for mixture lower than that for one of the components A	[66]
Ascorbic acid + trehalose, catechin + trehalose, gallic acid + trehalose, quercetin + trehalose	DPPH• decolorization	Antagonistic interaction, CI increasing with rising concentration of antioxidants. The role of trehalose unclear (it does not reduce DPPH•) T	[67]
α-Tocopherol + γ-oryzanol,	DPPH• decolorization	Synergy (20 + 200, 40 + 400, 60 + 600, 80 + 800, 1000 + 10,000 mg/kg) A	[68]
γ-Oryzanol + phytosterol in refined coconut oil	DPPH• decolorization	Additivity (20 + 200 mg/kg) Synergy (40 + 400, 60 + 600, 80 + 800, 1000 + 10,000 mg/kg) A	[68]
α-Tocopherol + phytosterol in refined coconut oil	DPPH• decolorization	Synergy (20 + 200, 40 + 400, 60 + 600, 80 + 800, 1000 + 10,000 mg/kg) A	[68]
α-Tocopherol + γ-oryzanol in purified rice bran oil	DPPH• decolorization	Antagonism (50 + 2000, 100 + 4000, 200 + 8000, 400 + 10,000, 500 + 12,000 mg/kg) A	[69]
α-Tocopherol + phytosterol in purified rice bran oil	DPPH• decolorization	Antagonism (50 + 2000, 100 + 4000, 200 + 8000, 400 + 10,000, 500 + 12,000 mg/kg) A	[69]
γ-Oryzanol + phytosterol in purified rice bran oil	DPPH• decolorization	Synergy (2000 + 2000, 4000 + 4000, 8000 + 8000, 10,000 + 10,000 and 12,000 + 12,000 mg/kg) A	[69]
Quercetin + esculetin, caffeoyl tartaric acid + gentisic acid, caffeoyl tartaric acid + esculetin, esculetin + gentisic acid, quercetin 3-O-arabinoside + gentisic acid, quercetin + gentisic acid, quercetin 3-O-arabinoside	DPPH• decolorization	Synergy (CI < 1) A	[70]
Quercetin + caffeoyl tartaric acid, quercetin 3-O-arabinoside + caffeoyl tartaric acid, rhamnetin + cyanidin 3-O-galactoside, quercetin + cyanidin 3-O-galactoside, quercetin 3-O-arabinoside + cyanidin 3-O-galactoside, quercetin 3-O-arabinoside + rhamnetin, cyanidin 3-O-galactoside + caffeoyl tartaric acid, cyanidin 3-O-galactoside + esculetin	DPPH• decolorization	Antagonism (CI > 1) A	[70]
Flavonoids	DPPH• decolorization	Antagonistic, additive or synergistc effects (from 11.9% to 121.7%)	[71]
S-Allyl-L-cysteine + catechin	DPPH• decolorization	Synergy (CI < 1) TK	[64]
S-Allyl-L-cysteine + quercetin		Synergy (CI < 1) TK
S-Allyl-L-cysteine + caffeic acid		Synergy (CI < 1) TK
S-Allyl-L-cysteine + ferulic acid		Synergy (CI < 1) TK
S-Allyl-L-cysteine + sinapic acid		Synergy (CI < 1) TK
S-Allyl-L-cysteine + 3,4-dihydroxybenzoic acid		Synergy (CI < 1) TK
Flavonoids	FRAP	Antagonistic, additive or synergistic effects (from 91.9.1% to 122.4%)	[71]
S-Allyl-L-cysteine + catechin	Ferricyanide reduction	Synergy (CI < 1) D	[64]
S-Allyl-L-cysteine + quercetin		Synergy (CI < 1) D
S-Allyl-L-cysteine + caffeic acid		Synergy (CI < 1) D
S-Allyl-L-cysteine + ferulic acid		Antagonism (CI > 1) D
S-Allyl-L-cysteine + sinapic acid		Synergy (CI < 1) D
S-Allyl-L-cysteine + 3,4-dihydroxybenzoic acid		Additivity (CI ≈ 1) D
Resveratrol + caffeic acid, equimolar	Ferricyanide reduction	Additivity (CI ≈ 1) N	[72]
Quercetin + resveratrol, equimolar Quercetin + caffeic acid, equimolar	Nonsite-specific hydroxyl radical-mediated 2-deoxy-D-ribose degradation, ferric ion reducing power	Synergy (CI < 1) N	[72]
Resveratrol + caffeic acid, equimolar	Nonsite-specific hydroxyl radical-mediated 2-deoxy-D-ribose degradation, FRAP, NO• scavenging	Synergy(CI < 1) N	[72]
Quercetin + resveratrol + caffeic acid, equimolar	Nonsite-specific hydroxyl radical-mediated 2-deoxy-D-ribose degradation, ferric reducing power, FRAP, NO• scavenging	Synergy(CI < 1) N	[72]
Resveratrol + caffeic acid, equimolar	Site-specific hydroxyl radical-mediated 2-deoxy-D-ribose degradation, DPPH• decolorization, ABTS• decolorization	Antagonism (CI > 1) N	[72]
Quercetin + resveratrol + caffeic acid, equimolar	Site-specific hydroxyl radical-mediated 2-deoxy-D-ribose degradation, DPPH• decolorization, ABTS• decolorization	Antagonism (CI > 1) N	[72]
Quercetin + α-tocopherol	Oxidation of methyl linoleate, reaction half-time	Synergy or additivity A2	[73]
Quercetin + astaxanthin	Oxidation of methyl linoleate, reaction half-time	Additivity A2	[43]
Quercetin + rutin	Oxidation of methyl linoleate, reaction half-time	Synergy A2	[73]
γ-Terpinene + PMHC or CAPE	Oxidation of styrene, lag time	Prolongation of the induction time by γ-terpinene, which did not show any induction time itself N	[74]
Quercetin + rutin	Peroxidation of sunflower oil, lag time	Additivity A2	[73]
RA + CHA, RA + EC, RA + AA, RA + GA, RA + PC, EC + PC, CHA + PC, CHA + EC, EC + AA, CHA + GA, GA + PC, PC + GA + AA + EC + CHA + RA PC + AA + EC + CHA + RA PC + GA + EC + CHA + RA PC + GA + AA + CHA + RA GA + AA + EC + CHA + RA PC + GA + AA + EC + RA PC + GA + AA + EC + CHA	Photochemical electrode, amperometric	Synergy or antagonism N	[75]
α-Tocopherol + γ-oryzanol, in refined coconut oil	Lag time (Rancimat)	Synergy (20 + 200, 40 + 400, 60 + 600, 80 + 800, 1000 + 10,000 mg/kg) A	[68]
α-Tocopherol + phytosterol in refined coconut oil	Lag time (Rancimat)	Antagonism (20 + 200 mg/kg) Synergy (40 + 400, 60 + 600, 80 + 800, 1000 + 10,000 mg/kg) A	[68]
γ-Oryzanol + phytosterol in refined coconut oil	Lag time (Rancimat)	Antagonism (20 + 200 mg/kg) Synergy (40 + 400, 60 + 600, 80 + 800, 1000 + 10,000 mg/kg) A	[68]
α-Tocopherol + γ-oryzanol in purified rice bran oil	Lag time (Rancimat)	Antagonism (50 + 2000, 100 + 4000, 200 + 8000, 400 + 10,000, 500 + 12,000 mg/kg) A	[69]
α-Tocopherol + phytosterol in purified rice bran oil	Lag time (Rancimat)	Antagonism (50 + 2000, 100 + 4000, 400 + 10,000, 500 + 12,000 mg/kg) or additivity (200 + 8000 mg/kg) A	[69]
γ-Oryzanol + phytosterol in purified rice bran oil	Lag time (Rancimat)	Synergy (2000 + 2000, 8000 + 8000, 10,000 + 10,000 and 12,000 + 12,000 mg/kg) Antagonism (4000 + 4000 mg/kg) A	[69]
Caffeoyl tartaric acid + gentisic acid, quercetin + quercetin 3-O-arabinoside, quercetin + caffeoyl tartaric acid;	Lard oxidation, lag time (Rancimat)	Synergy (CI < 1) A	[70]
α-Tocopherol + rosmarinic acid	AAPH-induced oxidation of linoleic acid, inhibition of peroxide formation	Synergy (CI < 1) TK	[76]
α-Tocopherol + rosmarinic acid	AAPH-induced oxidation of linoleic acid, inhibition of hexanal formation	Synergy (CI < 1) TK	[76]
α-Tocopherol + caffeic acid		Synergy (CI < 1) TK
α-Tocopherol + luteolin		Synergy (CI < 1) TK
α-Tocopherol + rosmarinic acid	AAPH-induced oxidation of linoleic acid, inhibition of (E)-2-heptanal formation	Synergy (CI < 1) TK	[76]
α-Tocopherol + caffeic acid		Additivity (CI ≈ 1) TK
α-Tocopherol + luteolin		Synergy (CI < 1) TK
α-Tocopherol + rosmarinic acid	AAPH-induced oxidation of linoleic acid, inhibition of (E)-2-nonenal formation	Synergy (CI < 1) TK	[76]
α-Tocopherol + caffeic acid		Synergy (CI < 1) TK
α-Tocopherol + luteolin		Synergy (CI < 1) TK
α-Tocopherol + rosmarinic acid	AAPH-induced oxidation of linoleic acid, inhibition of (E,E)-2,4-decadienal formation	Antagonism (CI > 1) TK	[76]
α-Tocopherol + caffeic acid		Additivity (CI ≈ 1) TK
α-Tocopherol + luteolin		Synergy (CI < 1) TK
EGCG + kaempferol	Cellular antioxidant activity		[77]
6:1.5		Synergy (CI < 1) AD
7:1.5		Antagonism (CI > 1) AD

Interaction parameter < 100%, antagonism; interaction parameter > 100%, synergy. AA, ascorbic acid; CAPE, caffeic acid phenethyl ester; CHA, catechinic acid; EC, epicatechin; GA, gallic acid; GSH, glutathione; PC, proanthocyanidins; PMHC, 2,2,5,7,8-pentamethyl-6-chromanol; RA, resveratrol. Green color indicates synergistic interactions, red color indicates antagonistic interactions; ^N, no statistics; ^A, ANOVA; ^A2; two-way ANOVA; ^D, Duncan’s multiple range tests; ^TK, Tukey–Kramer test.

Some unusually high synergistic effects reported for some ORAC combinations are surprising. The authors do not propose an explanation for such strong synergy. It can be suspected that the deviation from the linearity of the assay may be involved, since it is known that ORAC assay is linear only within a range of antioxidant concentrations and, in general, a quadratic regression equation is often used to describe the relationship between the antioxidant concentration and the area under fluorescence curve as a measure of antioxidant activity [41].

Studies of interactions between various antioxidants show that the type of interaction depends on many variables, as discussed below.

3.1. The Type of Interaction Depends on the Assay

The same mixture of compounds can show different types of interaction when assessed by different assays. A mixture of ascorbic acid, chlorogenic acid, and cysteine showed additivity in superoxide scavenging, antagonism in scavenging of the hydroxyl radical, and synergy in the scavenging of tert–butyl peroxide radical in an EPR assay [60]. Interaction of curcumin with (−)-epicatechin or with the epicatechin fraction from green tea showed synergy in the ABTS^• decolorization assay but antagonism in the inhibition of lipid peroxidation. In the ABTS^• decolorization assay, the strongest effect was observed with the equimolar mixture of pure compounds [28].

The mechanisms of various assays are different (Table 1), which may be one reason for the discrepancies in results on the interactions of antioxidants. However, comparison of interactions of S-allyl-L-cysteine with a range of phenolics showed different types of interactions (synergy vs. antagonism) in two assays (ABTS^• decolorization and DPPH^• decolorization) acting via the same mechanism (SPLET) [64] (Table 2).

Even within one method, classification of the type of interaction may depend on the method of prediction of the additive effect. A study of interactions between flavonoids (taxifolin, quercetin, rutin, and morin) and glutathione (GSH) found weak antagonistic effects (90–97% of the expected additive effect) when using the “traditional” method of analysis based on the sum of results for individual compounds and mostly weak synergistic effects (99–111% of the additive effect) when analyzing the data using Webb’s simulation. In the first mode of analysis, a weak concentration effect was observed in the taxifolin-GSH and rutin-GSH systems, the interaction effect decreasing with the decrease in the flavonoid/GSH ratio [20]. Authors of this study also compared the ABTS^• decolorization assay with another assay employing ABTS, in which lag time caused by antioxidants was measured after the addition of potassium persulfate to the solution of ABTS. In this method, strong synergistic effects were observed in the systems of taxifolin-GSH, quercetin-GSH, and rutin-GSH, while a weak antagonistic effect or no significant effects were found in the morin-GSH system. The authors concluded that the manifestation of the synergy strongly depends on the mode of inflow of the free radicals.

In studies of the interactions of various antioxidants with red cabbage extract, synergy was found for glutathione in the FRAP assay but not in the ABTS^• decolorization and ORAC, and for Trolox in the ORAC assay but not in ABTS^• decolorization and FRAP [78,79].

The reactivity of antioxidants for various indicators used in the antioxidant assays differs [80], and results may depend on the specific interactions with indicators rather than on interactions between antioxidants. Therefore, the use of more than one analytical approach should be recommended for the assessment of antioxidant synergy/antagonism.

3.2. The Type of Interaction Depends on the Identity of Interacting Compounds

The dependence of interactions on the classes of interacting compounds seems obvious, but even within the defined groups of similar compounds, the type of interaction varies for different compounds. In the DPPH^• decolorization assay of antioxidant activity, equimolar mixtures of gallic and protocatechuic acid, of gallic acid and chlorogenic acid, and of protocatechuic acid and chlorogenic acid showed higher antioxidant activity than sums of activities of individual components (synergistic interaction), while the antioxidant activity of a mixture of gallic acid and vanillic acid was lower than the sum of activities of individual components (antagonistic interaction) [52].

In the ORAC assay employing 0.625 µM final antioxidant concentrations in the reaction mixture, the ORAC values for a combination of two hydroxybenzoic acids were mainly synergistic and ranged from 70% (gallic acid + vanillic acid) to 336% (gentisic acid + syringic acid) of the value expected for additive interactions. Interactions between three (from 102 to 148% of the values expected for additivity) and four hydroxybenzoic acids (139–167% of the values expected for additivity) were mostly or totally synergistic [28].

Among combinations of two hydroxycinnamic acids, the interactions were mainly synergistic (from 98% of the expected value for p-coumaric acid + rosmarinic acid) to 411% of the expected value (p-coumaric acid + ferulic acid). Interactions between three hydroxycinnamic acids were mainly antagonistic (78–116% of respective values predicted for additivity) and interactions between four acids were antagonistic, additive, or synergistic (86–144% of the values predicted for additivity) [28].

There was no synergistic effect in the interactions between the selected phenolics (catechin, chlorogenic acid, cyanidin, cyanidin 3-glucoside, cyanidin 3-rutinoside, epicatechin, peonidin, peonidin 3-glucoside, quercetin, quercetin 3-glucoside, quercetin 3-galactoside, and quercetin 3-rutinoside) in the ABTS^• decolorization assay, and only additive effects were observed, although in some cases the effects were apparently more than additive (but the deviations from the additivity did not exceed several percent), and in some cases antagonistic effects occurred [81].

A photochemical electrode based on a g-C₃N₄/NiS/TiO₂ sensor demonstrated synergistic interactions in various binary mixtures between resveratrol, catechinic acid, epicatechin, proanthocyanidins, pyrogallol, ascorbic acid, and gallic acid, as well as in complex mixtures composed of five or six components. However, an antagonistic interaction was noted for a system of proanthocyanidins/gallic acid [75].

Catechin significantly decreased the rate of decolorization of malvidin 3-glucoside and peonidin 3-glucoside but not delphinidin 3-glucoside, petunidin 3-glucoside, or cyanidin 3-glucoside during AAPH-induced peroxidation of linoleic acid and a considerably higher prolonging effect on the lag time of oxygen consumption was observed for malvidin 3-glucoside and peonidin 3-glucoside than for three other anthocyanins [82].

In a study of antioxidant interaction in the off-odor formation during oxidation of linoleic acid, the type of interaction changed from synergy to additivity and even to antagonism depending on the aldehyde product analyzed [76].

3.3. The Type of Interaction Depends on the Absolute Concentrations and the Concentration Ratio of Interacting Compounds

In studies of interactions between hydroxybenzoic acids (protocatechuic, gentisic, gallic, vanillic, and syringic acids) at a 3.2 µM concentration (final concentration in the reaction mixture) in the FRAP assay, all mixtures containing gentisic acid showed a synergistic effect (128–189% of the values expected for additivity). The mixture of protocatechuic and syringic acid showed an additive effect, while all others exerted an antagonistic effect (down to 42% of the values expected for additivity). At a concentration of 16.1 µM, a synergistic effect was observed only for the mixture of gentisic + syringic acids, while at 32.3 µM, the synergistic effect disappeared. Among ternary mixtures, the mixture of protocatechuic + gentisic + syringic acids showed the greatest synergistic effect (274% of the value expected for additivity at the 3.2 µM concentration), but this effect decreased at the concentration of 16.1% (107% of the expected value) and vanished at the concentration of 32.3 µM (99% of the expected value). Most of the ternary combinations of hydroxycinnamic acids showed antagonistic interactions; one of the few exceptions was the mixture of p-coumaric acid, caffeic acid, and rosmarinic acid, showing synergistic interactions (227% of the value expected for additivity at 3.2 µM concentration, 148% of the expected value at 16.1 µM concentration, and 111% of the expected value at 32.3 µM concentration). For combinations of four hydroxycinnamic acids, antagonistic interactions predominated at the concentrations of 3.2 and 16.1 µM, with additive or weakly synergistic interactions at 32.3 µM [52].

A FRAP study of the interactions between flavonoids in binary systems at various proportions of components, from 3:1 to 1:3, showed again the dependence of the effects on the antioxidant concentration ratio. In the catechin + quercetin and catechin + quercetin-3-β-glucoside systems, synergy was observed; the magnitude of the synergistic effect increased with a decreasing proportion of catechin (from 28% to 37% and from 33% to 51%, respectively). However, no such regularity was observed in other binary systems [71].

In the DPPH^• decolorization assay, quercetin showed a synergistic interaction with resveratrol in a molar ratio of 1:1, 2:1, and 3:1, but antagonistic interaction in ratios of 1:2 and 1:3. Hydroquinone showed a synergistic effect with kaempferol over a molar ratio of 1:3–3:1, with a maximal effect at the 2:1 ratio. The interaction of hydroquinone with resveratrol was weakly synergistic over the concentration range of 1:2 to 3:1. The interaction of kaempferol with resveratrol was weakly antagonistic over a ratio of 1:3. In contrast, the synergistic effects of interactions of rutin with resveratrol increased progressively with the increase in the rutin/resveratrol molar ratio in the range of 1:3 to 3:1, and the effect of quercetin/resveratrol rose progressively from antagonism (quercetin/resveratrol molar ratio of 1:3 and 1:2) to synergy at increasing quercetin/resveratrol ratios, up to 3:1 [66].

In a study of interactions between trehalose and ascorbic acid, catechin, gallic acid, and quercetin, antagonistic interactions were observed, with the interaction coefficient decreasing as antioxidant concentrations decreased [67]. However, interpretation of these results is not straightforward since trehalose did not reduce DPPH^• at all, so the effect cannot be referred to as an interaction between antioxidants, but was rather conditioned by other mechanism(s).

Binary antioxidant formulations: tert–butyl hydroxyquinone (TBHQ):butyl-4- hydroxytoluene (BHA), tert–butyl hydroxyquinone (TBHQ):propyl gallate, and TBHQ: pyrogallol showed the best synergistic effect at 2:1, 1:1, 2:1 weight ratios, respectively, in both distilled soybean-oil- and distilled poultry-fat-based biodiesel [83].

3.4. The Type of Interaction Depends on the Reaction Medium

The type of interaction between antioxidants obviously depends on the reaction medium. Viscosity affects the reaction rates of antioxidants [84]. A change in pH and decrease in medium polarity affect the ionization state and solubility, partitioning between phases; thus, interactions between antioxidants promote aggregation of less polar antioxidants. Redox potentials of the reagents, reaction kinetics, and predominant antioxidant mechanisms may all change with pH. The appearance of amphiphilic structures leads to partitioning of more hydrophobic antioxidants, which changes their local concentrations and their interactions, with hydrophilic antioxidants remaining in the aqueous phase, complicating reaction kinetics with respect to a homogeneous system [85,86,87,88]. A “polar paradox” observed in such systems consists in a tendency to concentrate non-polar antioxidants in the lipid environment, so if the oxidation process takes place in the hydrophobic phase, the apparent activity of non-polar antioxidants is higher than that of polar antioxidants [89]. Micellar charge will also affect the distribution of ionic antioxidants [90].

α-Tocopherol and ϒ-oryzanol showed mostly antagonistic interactions in rice bran oil [69] but a synergistic interaction in refined coconut oil [70]. It supports the view that the same combination of minor constituents may show diverse antioxidant interactions in different lipid environments.

The interaction between resveratrol and a tocopherol analogue, 2,2,5,7,8-pentamethyl-6-chromanol (PMHC), in the prevention of oxidation of methyl linoleate was additive in Triton X-100 [58] and in SDS micelles but synergistic in CTAB micelles [91] and a liposomal system. In DMPC liposomes, it depended on pH in the range of 4–8, with maximal synergy observed at pH 7. This effect was attributed to the pH dependence of the localization of resveratrol within the lipid region, affecting the probability of reaction of this compound with the tocopheroxyl radical [58]. The interaction of resveratrol with PMHC in the inhibition of lipid peroxidation with PMHC was additive in SDS micelles, and synergy was observed in CTAB micelles [92]. Further examples of the dependencies shown above can be found in Table 2 and Table 3.

The dependence of antioxidant interaction on the medium is very important and should be considered in attempts to extrapolate results of in vitro experiments to in vivo situations or food systems.

4. Mechanisms of Interactions Between Antioxidants

Several mechanisms have been proposed to account for the non-additivity of antioxidant interactions in vitro; first of all, the regeneration of antioxidants.

4.1. Regeneration of Antioxidants

In an assay based on a reduction of a free radical of an indicator like DPPH^• or ABTS^• decolorization assay, an antioxidant A reacts with the free radical (e.g., ABTS^•), reducing it to the non-radical indicator molecule (ABTS) and forming the antioxidant free radical A^•

A + ABTS^• → A^• + ABTS (5)

The antioxidant free radical can react with a second radical of the indicator, reducing it and forming a two-electron oxidation product of the antioxidant A_ox

A^• + ABTS^• → A_ox + ABTS (6)

Alternatively, A^• can be reduced by another antioxidant, forming a free radical of another antioxidant B, regenerating the initial, reduced form of the antioxidant [93,94]:

A^• + B →A + B^• (7)

In the ORAC assay, the situation is somewhat more complicated: antioxidants can react with peroxyl radicals, preventing oxidation of the indicator (usually fluorescein), or regenerate the indicator from its radical formed upon oxidation by peroxyl or alkoxyl radicals [40,41,95].

Most often, the synergistic interaction between antioxidants has been attributed to regeneration of the antioxidant radical or fully oxidized antioxidant formed in the assay by another antioxidant of lower redox potential. The classic example is the regeneration of tocopherol from the tocopheryl radical by ascorbic acid [96,97]. Ascorbic acid was postulated to regenerate the tocopheryl radical formed in reactions of tocopherol with peroxyl radicals of lipids undergoing peroxidation. This regeneration is thermodynamically possible, taking into account the standard redox potentials of the tocopheryl radical/tocopherol redox pair (500 mV) and semidehydroascorbyl radical/ascorbate anion redox pair (282 mV) [98]. In a homogeneous medium containing both tocopherol and ascorbate, ascorbate was utilized first [97]. In a heterogenous system, this synergism may be even more important. α-Tocopherol is lipid-soluble, but its hydroxyl group and phytol chain give it amphiphilic properties. Therefore, α-tocopherol partitions at the interface of cell membranes to protect against lipid oxidation. When free radicals are present within the membrane, α-tocopherol is oxidized to form a more polar α-tocopheroxyl radical, which can interact with water-soluble ascorbic acid at the membrane surface and initiate the electron transfer required to regenerate α-tocopherol back to its original form.

ROO^• + Tocopherol → ROOH + Tocopherol^• (8)

Tocopherol^• + Ascorbate⁻ → Tocopherol + Ascorbate^•− (9)

where ROO^• is a lipid peroxyl radical and ROOH is lipid hydroperoxide [97].

The ascorbate radical is relatively stable and can disproportionate, forming ascorbate and dehydroascorbate [99].

Ascorbate^•− + Ascorbate^•− → Ascorbate⁻ + Dehydroascorbate⁻ (10)

In the egg phosphatidylcholine–dicetylphosphatidylcholine liposome system, lipid peroxidation was significantly retarded in the presence of tocopherol + ascorbate compared with tocopherol alone [100]. Regeneration of tocopherol was also observed in membrane models and in biological membranes [101]. Glutathione did not affect the protective effect of tocopherol significantly [100].

The property to regenerate tocopherol from its radical was postulated to be shared by other compounds. Ubiquinol and phenolic compounds such as epigallocatechin gallate (EGCG), gallic acid, and epicatechin were also reported to regenerate α-tocopherol from its radical [102,103]:

Ubiquinol + Tocopherol^• → Ubiquinol^• + Tocopherol (11)

Phenolic + Tocopherol^• → Phenolic^• + Tocopherol (12)

A synergistic interaction was observed in a system of alpha-tocopherol + green tea catechins (epigallocatechin gallate, epigallocatechin, or epicatechin gallate (ECG)) + ascorbic acid in the inhibition of linoleic acid in SDS micelles. While tocopherol and catechins induced a lag period, ascorbic acid alone did not delay the onset of substrate oxidation. In the system containing tocopherol, a catechin, and ascorbic acid, ascorbic acid was consumed first, followed by catechin and tocopherol as the last compound. Addition of ascorbic acid to ECG solution changed the EPR signal of the ECG radical into the signal of the ascorbyl radical, pointing to the regeneration of the alpha-tocopherol radical as the mechanism for the synergistic action of catechins and ascorbate [59].

Autoxidation of several model substrates (stripped sunflower oil, squalene, and styrene) was inhibited by PMHC and caffeic acid phenethyl ester (CAPE). With all substrates, γ-terpinene acted synergistically. The combination of either PMHC or CAPE with γ–terpinene did not significantly change the slope of the inhibited period of styrene oxidation compared to that produced by each of the phenols alone, but it extended its duration in a dose-dependent fashion while not retarding the onset of styrene oxidation itself [74]. Similar results were obtained in studies of oxidation of stripped sunflower oil at 130 °C, where γ–terpinene prolonged the antioxidant activity of α-tocopherol and PMHC [104].

The synergy of γ-terpinene with PMHC or CAPE was ascribed to the regeneration of the chain-breaking antioxidants PMHC and CAPE from their radicals to original catechols [74]. Similarly, pyrogallol, being the more effective antioxidant, readily donates a hydrogen atom from its hydroxyl group to fatty acid free radicals, thereby forming an antioxidant radical. TBHQ then transfers hydrogen to the antioxidant radical to regenerate it back to pyrogallol

ROO^• + Pyrogallol → ROOH + Pyrogallol^• (13)

Pyrogallol^• + TBHQ → Pyrogallol + TBHQ^• (14)

In the process, TBHQ is converted to a radical that can form stable products with other free radicals, and this effect, together with the interaction and regeneration of pyrogallol, represents an effective synergistic effect between the two antioxidants. Analysis of distilled soybean-oil-based biodiesel stored for 3 months with 2:1 TBHQ:pyrogallol showed the total amount of pyrogallol close to its original value [83].

The synergism between β-carotene and astaxanthin or between lycopene and astaxanthin, and the lack of synergism between zeaxanthin and astaxanthin in the inhibition of lipid peroxidation induced by the hydrophobic initiator AMVN, has been explained by the regeneration of astaxanthin at the expense of β-carotene or lycopene. According to this proposal, astaxanthin, located at the lipid–water interface, acts as a radical bridge channeling the electron from the more reducing carotenoid in the inner part of the membrane to the interface, where the radicals with high dipole moment are concentrated. Zeaxanthin and astaxanthin do not act synergistically, as they both have similar spatial distribution in the membrane [56].

ROO^• + Astaxanthin → ROOH + Astaxanthin^• (15)

Astaxanthin^• + Lycopene → Astaxanthin + Lycopene^• (16)

Resveratrol was reported to be neither an outstanding antioxidant nor a co-antioxidant in homogeneous solutions [105] at the lipid–water interface. Resveratrol scavenges LOO^• within the liposomal membrane; thus, resveratrol belongs to the same class of lipophilic antioxidants as α-tocopherol [106].

The mere regeneration of one antioxidant by another does not change the amount of free radicals in the system and cannot explain the synergy of antioxidant action. However, it can lead to the effect of synergy, taking into account the broader context of antioxidant reactivity.

• (i)The second antioxidant is not reactive with the probe but reacts with A^•. This may seem improbable since the thermodynamic condition for A to react with the free radical indicator is that the redox potential of the redox system (A^•/A) must be lower than the redox potential of the redox system (indicator radical/reduced indicator). To be able to regenerate A, the redox potential of the (B^•/B) system must be lower than the redox potential of the (A^•/A) system. If so, why should it not reduce the indicator radical itself? In principle, it should, but apart from the thermodynamic condition, kinetic factors (e.g., caused by a steric obstacle) may cause such a reaction to be slow or impossible [80]. If the reaction is slow, it may not be completed within the limited time of the assay. Regeneration of the more reactive antioxidant A allows it to reduce more indicator molecules within the time frame of the assay (Figure 1A).
• (ii)The free radical of an antioxidant such as Trolox may react with the free radical of an indicator. However, for other antioxidants, this radical may not be as reactive as the native antioxidant. In such a case, regeneration of A from A^• will increase the reactivity of A. This mechanism will be important if antioxidant B is less reactive for the indicator than A; otherwise, it would hardly affect the overall activity of antioxidants in the system (Figure 1B).

Figure 1.

Mechanisms of non-additive interactions between antioxidants (presented in the literature and hypothetical). Regeneration of antioxidant A from its radical may increase the antioxidant activity within the reaction time frame if another antioxidant B reacts with the indicator (Ind) at a slower rate (A) or if antioxidant A regenerates B from its radical that has low reactivity for the indicator (B). Ind^•, free radical indicator, e.g., ABTS^• or DPPH^•; IndH, reduced indicator. (C), Change of reaction mechanism: antioxidants such as γ-terpinene (T) change the propagating radical from LOO^• to HOO^•, more susceptible to termination [107]. (D) Formation of an adduct AB, less reactive than the still-reactive further oxidation products A_fop of A [83]. (E) Non-covalent or covalent interactions with another antioxidant may decrease or increase the reactivity of antioxidants [60]. (F) Metal chelation may affect the reactivity of antioxidants; e.g., flavonoids [108,109]. Red, oxidant; Green, antioxidant; Blue, less reactive oxidant or antioxidant.

The mechanism (ii) has been considered in an attempt to explain the differences in the extent of cooperation of catechin with different anthocyanins. The spin densities on the phenolic O atoms estimated by a semiempirical quantum chemical method differed for various anthocyanins. The radicals generated by delphinidin 3-glucoside, petunidin 3-glucoside, and cyanidin 3-glucoside (first class), having an o-dihydroxy structure on the B ring, show a low spin density (3–4%) on the phenolic O atoms, while those generated by malvidin 3-glucoside, peonidin 3-glucoside, and pelargonidin 3-glucoside (second class) are characterized by a spin density on phenolic O atoms of the order of 24% [82]:

ROO^• + Anthocyanin → ROOH + Anthocyanin^•

Anthocyanin^• + Catechin → Anthocyanin + Catechin^•

However, in both cases, considering reactions in a real system, the synergy found in an in vitro assay may represent an experimental artefact since, over a longer time, the slower antioxidant would have a chance to react as well, especially when exposed to a continuous flux of reactive oxygen species (ROS).

4.2. Change of Reaction Mechanism

One antioxidant can change the reaction mechanism in a way favorable for another antioxidant. Another effect was proposed to explain the synergistic effect of γ-terpinene in the peroxidation of lipids. According to this proposal, terpinene changes the propagation chain-carrier from the peroxyl radicals ROO^• to hydroperoxyl radicals HOO^• (Figure 1C). γ-Terpinene is rapidly attacked by ROO^• and releases HOO^• [104].

γ-Terpinene + ROO^• → γ-Terpinene^• + ROOH

γ-Terpinene^• + O₂ → HOO^• + p-Cymene

Hydroperoxyl radicals can both propagate the oxidation and be quenched by another HOO^• or by ROO^• (self-termination or cross-termination). Since the self-termination of HOO^• and its cross-termination with ROO^• is much faster than the self-termination of ROO^•, the overall termination efficiency would increase in the presence of γ-terpinene [104,107].

4.3. Formation of Antioxidant Hetero- and Homodimers, and Other Reactive Products

Generally, the set of reactions occurring during assays of antioxidant activity is much broader than those symbolized in Equations (5)–(7). They may include recombination between antioxidant radicals and further reactions of oxidized antioxidants. For example, GSH is oxidized to glutathione disulfide (GSSG), but GSSG can still react with ABTS^• to form higher oxidation products [20,32]. Reaction of the GSH radical with a radical of another antioxidant leads to the formation of a glutathione conjugate. If the conjugate is less reactive than GSSG, this reaction will contribute to an antagonistic interaction between both antioxidants (Figure 1D); if it is more reactive, it may lead to a synergistic interaction. Analogous situations can occur with other pairs of antioxidants. Some authors classify such interactions as “apparent synergy” [110].

Formation of dimers between antioxidant radicals may produce new antioxidant species that contain two phenols, which in effect are stronger or weaker antioxidants than the parent antioxidants (Figure 1E). Examples of such dimers that are stronger antioxidants are shown in Figure 2.

Figure 2.

A heterodimer formed in the reaction of TBHQ and pyrogallol [83] (A), and 2,2-dimethyl-2,3-dihydro-benzo(b)furan, a reactive product formed during oxidation of TBHQ (B) [111].

Flavonoids lacking the catechol group in the B ring showed antagonistic behavior with GSH. Myricetin displayed an additive effect, while quercetin, fisetin, luteolin, luteolin-7-O-glucoside, taxifolin, and (+)-catechin demonstrated synergistic actions. Quinones formed by oxidation of phenol groups may react with GSH, forming adducts [112], which may affect DPPH^• scavenging. Among the flavonoids tested, those having a catechol group in the B ring, namely quercetin, fisetin, luteolin, luteolin-7-O-glucoside, taxifolin, and (+)-catechin, showed synergistic effects with glutathione. An exception was noted with quercetagetin, which has one additional hydroxyl group at C6 of the A ring compared to quercetin. Adducts formed at C2′ and C5′ of the B ring seem to be more important for the DPPH^• reduction than adducts formed at C6 and C8 of the A ring [27].

Antioxidant homodimers, higher oligomers, or further degradation products may also be formed, and if they retain the reactivity with the oxidant (or if their reactivity is even enhanced), this effect may contribute to synergy between antioxidants when another antioxidant facilitates their formation (e.g., when a dimer-forming antioxidant regenerates another antioxidant). Resveratrol radicals formed in reaction with peroxyl radicals during lipid peroxidation may form dimers and further cyclization products, with recovered hydroxyl groups able to trap peroxyl radicals. The formation of such dimers is facilitated in micellar and liposomal systems due to the increased concentration of resveratrol at the lipid–water interface at pH 6–7 and may be responsible for increased antioxidant activity and synergistic effects of resveratrol [106]. A benzo(b)furan derivative formed during oxidation of TBHQ, 2,2-dimethyl-2,3-dihydro-benzo(b)furan, also shows a strong antioxidant activity [111]. Products of melatonin oxidation retain the antioxidant activity, forming a “radical scavenging cascade” [113,114]. Products of the reaction of chrysin with ABTS^• have an even higher reactivity for ABTS^• than the parent compound [115].

Sugars may cause a reduction in the antioxidant activity of the polyphenols. Condensation reactions can occur between sugar molecules and polyphenols; in consequence, glycosides like pentagalloylglucose and tetragalloylglucose are likely to be formed. However, sucrose molecules can also interact with oxidized phenolic compounds; as a result, reduced forms of phenolic may be formed, resulting in an increase in antioxidant activity [116].

4.4. Non-Covalent Interactions Between Antioxidants

Formation of intermolecular hydrogen bonds between antioxidant molecules may be an important factor. While compounds that are non-hydrogen-bonded (free) are expected to possess full activity in electron or hydrogen atom transfer mechanisms, this activity may be diminished by hydrogen bonding and modified by other non-covalent interactions [117,118,119]. When both carboxyl groups and phenolic hydroxyl groups are present in one system, possible bonding could occur between the two kinds of groups, which may account for the antagonistic interactions of gallic acid with proanthocyanidins [75]. H-bonding between flavonoids may result in a decrease in the availability of -OH groups and thus a decrease in the antioxidant activity [60].

4.5. Metal Chelation

Many antioxidants, including flavonoids, have the ability to chelate metal ions, which catalyze various oxidation reactions but may also have antioxidant properties in their lower oxidation states. Metal ions, especially ferrous or ferric ions, may be present as trace contaminants of reagents, e.g., buffer components or as components of the sample studied [120,121]. Antioxidants may bind these ions, reduce ferric ions, or both bind and reduce ions at higher oxidation states. Binding and reduction of metal ions may diminish the intrinsic antioxidant activity of organic compounds, but metal ions may increase the activity of the complex in comparison with the parent compound (Figure 1F). Both these effects can contribute to the hypo- or hyperadditivity of antioxidant mixtures [17]. Complexation of flavanols with metal ions mainly decreased, but in some cases increased their reactivity in the ABTS^• decolorization assay [108]. Metal complexes of flavonoids such as quercetin, rutin, galangin, and catechin had better activity in the DPPH^• decolorization assay than the parent flavonoids [109]. It could be expected that metal chelation can especially affect the FRAP assay, based on Fe(III) reduction. However, Fe(III) is present in high excess over antioxidants in this assay, and minute amounts that may be present in the analyzed sample do not interfere with the assay [37].

5. Interactions of Antioxidants with Proteins

Antioxidants may interact with proteins in food and with blood plasma proteins, forming non-covalent complexes or covalent conjugates, which may affect antioxidant reactivity and the antioxidant activities of proteins [17,122,123,124,125,126]. Complexation with proteins may protect the antioxidant activity of flavonoids in the digestive tract [125]. Often, such interactions enhance the antioxidant activity of proteins but decrease the antioxidant activity of small-molecular-weight antioxidants, although exceptions to this rule have been reported [127]. Binding of a small-molecular-weight antioxidant may induce conformational changes of a protein. Studies of the interactions of EGCG and caffeic acid with whey proteins demonstrated both synergistic and antagonistic effects. The former were explained by protein conformational changes that allow free radicals to access nucleophilic centers previously inaccessible because of steric hindrance, while the latter were ascribed to H-bond formation between protein groups and the hydroxyl groups of phenolic compounds, thereby reducing the number of available hydroxyl groups and altering the ability to donate electrons [125,127,128,129]. Digestion of protein complexes with small-molecular-weight antioxidants can lead to a release of the latter from the complexes. Formation of complexes between rice proteins and phenolic acids increased the antioxidant activity of the proteins but decreased their digestibility [130]. Chlorogenic acid showed synergy with wheat proteins and casein, especially after simulated digestion. It has also been hypothesized that polyphenol–protein systems can clear free radicals by forming more stable reaction products, thereby terminating the free radical chain reaction [126].

6. Interactions Between Antioxidants in the Folin–Ciocalteu Phenol Assay

The Folin–Ciocalteu assay for phenolic compounds is also a redox reaction of a complicated mechanism [131,132].

In the assay of total phenolics employing the Folin–Ciocalteu reagent, additivity was observed in the combination of extracts of different spices and pure phenolics with synthetic antioxidants, although in some cases a synergistic effect was evident [133]. When studying legume/non-legume/fruit food combinations, mostly antagonistic interactions were observed [134]. For mixtures of green tea (Camellia sinensis), honey, and lemon (Citrus limonum) extract, synergistic effects were found (144.0–181.3% of values expected for additivity) [135].

In the mixtures of spirulina (Arthrospira) with apple (Malus domestica) juice and with Japanese quince (Chaenomeles japonica) syrup, the measured phenolic content was 77% and 74%, respectively, of the sum of those measured for single components [136].

Such observation casts doubt on the validity of the assay; due to interactions between the components of analyzed samples, its results may not reflect the real contents of these compounds.

7. Antioxidant Interactions in Complex Food Products

Antioxidant interactions between food products and extracts have also been the subject of numerous studies. Of course, in these studies, the mechanisms underlying these interactions are more difficult, if possible at all, to assess.

A synergistic interaction of green tea with honey and lemon extract was demonstrated in the ABTS^• decolorization assay [134]. Synergistic interactions were found between sweet potato (Ipomoea batatas) extracts, tea polyphenols, and kudzu vine (Pueraria) flavonoids by DPPH^• decolorization and FRAP assays [137].

Extracts of leaves of ivy gourd (Coccinia grandis) and sambong (Blumea balsamifera) showed both synergistic and antagonistic interactions in the ABTS^• decolorization assay, depending on the ratio of the extracts, and mostly synergistic interactions in the DPPH^• decolorization assay [138].

Antagonistic or synergistic interactions between ethanolic extracts of leaves of mango (Mangifera indica) and the aerial parts of asthma weed (Euphorbia hirta) were observed, depending on the ratio of the extracts, in the FRAP assay [139].

Interactions between herbal components of a TC-16 polyherbal formulation comprising turmeric (Curcuma longa), Bentong ginger (Zingiber officinale var. Bentong), black pepper (Piper nigrum), and calamondin (Citrofortunella macrocarpa) with giant honeybee (Apis dorsata) honey were antagonistic in the ABTS^• decolorization, DPPH^• decolorization, and FRAP assays and synergistic in ORAC and beta-carotene bleaching assays [140].

In the DPPH^• decolorization assay, strongly synergistic interactions were revealed between methanolic extracts of edible portions of khejri desert legumes, khejri (Prosopis cineraria) and gum Arabic tree (Acacia senegal), non-legumes (fragrant manjack Cordia dichotoma and karira Capparis decidua), and mango (experimental EC₅₀ values < expected EC₅₀ values). In the FRAP assay, mostly synergistic but also additive and antagonistic interactions were found [134].

Interaction of spirulina with apple juice was additive in the FRAP assay and antagonistic in the ABTS^• decolorization assay; interaction of spirulina and Japanese quince syrup was synergistic in the FRAP assay and additive in the ABTS^• decolorization assay; interaction between spirulina and cranberry (Vaccinium macrocarpon) syrup was synergistic in the FRAP assay and antagonistic in the ABTS^• decolorization assay [136]. The latter study also demonstrated considerable differences between interactions in initial food products and their bioavailable fractions.

Interactions between essential oils of bay laurel (Laurus nobilis), Spanish lavender (Lavandula stoechas), and pennyroyal (Mentha pulegium) were mainly synergistic but additive or antagonistic in some cases, depending on the proportions of the oils and method of assay (ABTS^• decolorization, DPPH^• decolorization and CUPRAC) [141].

Mostly synergistic effects were found in the interaction between extracts of peel of bitter orange (Citrus aurantium) and aerial parts of field wormwood (Artemisia campestris) and white wormwood (Artemisia herba alba) in the ABTS^• decolorization and FRAP assays [142].

Analysis of interactions between various food products using the QUENCHER procedure [143] and ABTS^• decolorization assay demonstrated antagonistic, additive, and synergistic interactions between various products and, often, changes in the type of interaction after simulated gastric, intestinal, and colonic digestion (exemplary results shown in Table 4 and Table 5). Some of the observed antagonistic interactions can be due to the binding of polyphenols to milk proteins [144].

Table 4.

Exemplary quantitative results of interactions between food products in vitro.

System	Assay	Interaction Parameter = 100% × [(Effect Measured)/(Effect Predicted as the Arithmetic Sum of Individual Antioxidant Activities)]; Type of Interaction		Reference
32% fresh spirulina/68% apple juice	ABTS• decolorization	60	Ant N10	[136]
32% fresh spirulina, 43% Japanese quince juice, 25% sugar	ABTS• decolorization	97	Add N10	[136]
32% fresh spirulina, 43% Japanese quince juice, 25% sugar	ABTS• decolorization	97	Add N10	[136]
Jute + blackjack 3:7	ABTS• decolorization	142.9	Syn N10	[145]
Jute + blackjack 5:5	ABTS• decolorization	104.0	Add N10	[145]
Jute + blackjack 7:3	ABTS• decolorization	100.9	Add N10	[145]
Breakfast cereals + strawberry	Q/ABTS• decolorization	158.9	Syn T	[144]
		BF 92.4	Add T
Breakfast cereals + blueberry	Q/ABTS• decolorization	128.7	Syn T	[144]
Breakfast cereals + black grape	Q/ABTS• decolorization	160.5	Syn T	[144]
Breakfast cereals + hazelnut	Q/ABTS• decolorization	153.6	Syn T	[144]
		BF 40.7	Ant T
Breakfast cereals + chia seed	Q/ABTS• decolorization	109.4	Add T	[144]
Breakfast cereals + flaxseed	Q/ABTS• decolorization	101.9	Add T	[144]
		BF 88.1	Ant T
Breakfast cereals + yoghurt	Q/ABTS• decolorization	118.1	Syn T	[144]
Breakfast cereals + milk	Q/ABTS• decolorization	35.3	Ant T	[144]
		BF 118.8	Syn T
White wheat bread + green tea extract	Q/ABTS• decolorization	139.1	Syn T	[144]
		BF 38.9	Ant T
White wheat bread + black tea extract	Q/ABTS• decolorization	173.1	Syn T	[144]
		BF 123.7	Syn T
White wheat bread + flaxseed	Q/ABTS• decolorization	81.7	Ant T	[144]
		BF 72.4	Ant T
White wheat bread + chia seed	Q/ABTS• decolorization	67.9	Ant T	[144]
White wheat bread + sesame	Q/ABTS• decolorization	72.5	Ant T	[144]
White wheat bread + cheese	Q/ABTS• decolorization	91.0	Add T	[144]
White wheat bread + milk	Q/ABTS• decolorization	73.5	Ant T	[144]
Milk + strawberry	Q/ABTS• decolorization	73.6	Ant T	[144]
		BF 135.2	Syn T
Milk + blueberry	Q/ABTS• decolorization	70.6	Ant T	[144]
Milk + black grape	Q/ABTS• decolorization	81.5	Ant T	[144]
Yoghurt + strawberry	Q/ABTS• decolorization	118.3	Syn T	[144]
Yoghurt + blueberry	Q/ABTS• decolorization	156.5	Syn T	[144]
Yoghurt + black grape	Q/ABTS• decolorization	160.9	Syn T	[144]
Milk + chia seed	Q/ABTS• decolorization	43.7	Ant T	[144]
Milk + flaxseed	Q/ABTS• decolorization	54.7	Ant T	[144]
		BF 45.3	Ant T
Yoghurt + flaxseed	Q/ABTS• decolorization	166.4	Syn T	[144]
Yoghurt + chia seed	Q/ABTS• decolorization	100.2	Add T	[144]
Milk + green tea extract	Q/ABTS• decolorization	89.0	Ant T	[144]
		BF 40.8	Ant T
Milk + black tea extract	Q/ABTS• decolorization	234.4	Syn T	[144]
		BF 65.2	Ant T
Milk + espresso	Q/ABTS• decolorization	162.6	Syn T	[144]
		BF 87.7	Ant T
Jambon + white wheat bread	Q/ABTS• decolorization	83.8	Ant T	[144]
		BF 112.4	Syn T
Jambon + lettuce	Q/ABTS• decolorization	77.6	Ant T	[144]
Jambon + tomato	Q/ABTS• decolorization	76.4	Ant T	[144]
Jambon + cheese	Q/ABTS• decolorization	146.2	Syn T	[144]
		BF 99.9	Add T
Lettuce + tomato	Q/ABTS• decolorization	135.6	Syn T	[144]
Lettuce + cabbage	Q/ABTS• decolorization	113.0	Syn T	[144]
Tomato + white wheat bread	Q/ABTS• decolorization	119.2	Syn T	[144]
Tomato + cheese	Q/ABTS• decolorization	119.6	Syn T	[144]
Wine + tomato	Q/ABTS• decolorization	117.3	Syn T	[144]
Wine + white wheat bread	Q/ABTS• decolorization	114.3	Syn T	[144]
		BF 75.2	Ant T
Wine + cheese	Q/ABTS• decolorization	122.2	Syn T	[144]
		BF 94.9	Add T
Wine + jambon	Q/ABTS• decolorization	142.1	Syn T	[144]
		BF 124.9	Add T
Jute + blackjack 3:7	DPPH• decolorization	113.0	Syn N10	[145]
Jute + blackjack 5:5	DPPH• decolorization	109.2	Add N10	[145]
Jute + blackjack 7:3	DPPH• decolorization	101.9	Add N10	[145]
Jute + blackjack 3:7	FRAP	89.4	Ant N10	[145]
Jute + blackjack 5:5	FRAP	83.9	Ant N10	[145]
Jute + blackjack 7:3	FRAP	82.0	Ant N10	[145]
Ethanolic extracts of leaves of Mangifera indica + extracts of the aerial parts of Euphorbia hirta	FRAP			[139]
1:3		94.2	Add N10
1:1		78.8	Ant N10
Green tea extract + honey	FRAP	121.7	Syn A	[135]
Green tea extract + honey + lemon extract	FRAP	180.0	Syn A	[135]
32% fresh spirulina/68% apple juice	FRAP	102	Add N10	[136]
32% fresh spirulina, 43% Japanese quince juice, 25% sugar	FRAP	131	Syn N10	[136]
32% fresh spirulina/68% apple juice	Total phenolics	77	Ant N10	[136]
32% fresh spirulina, 43% Japanese quince juice, 25% sugar	Total phenolics	74	Ant N10	[136]

Interaction parameter < 100%, antagonism; interaction parameter > 100%, synergy. BF, bioaccessible fraction; Q, quencher approach. Add, additivity; Ant, antagonism; Syn, synergy; ^N5, criterion of 5% difference from 100%; ^N10, criterion of 10% difference from 100%. ^A, ANOVA; ^T, Student’s “t” test.

Table 5.

Exemplary qualitative results of interactions between food products in vitro.

System	Assay	Effects	Reference
Curcumin + epicatechin fraction of green tea	ABTS• decolorization	Synergy (CI 0.80–0.91) A	[28]
TEMPOL + red cabbage extract	ABTS• decolorization	Antagonism T	[78]
Trolox + red cabbage extract	ABTS• decolorization	Additivity T	[78]
Lavandula stoechas essential oil + Mentha pulegium essential oil, 1:2, 1:1.2:1	ABTS• decolorization	Antagonism (1:2, 1:1) A Synergy (2:1)	[141]
Laurus nobilis essential oil + Lavandula stoechas essential oil, 1:2, 1:1.2:1	ABTS• decolorization	Synergy (CI 0.15–0.54) A	[141]
Laurus nobilis essential oil + Mentha pulegium essential oil, 1:2, 1:1.2:1	ABTS• decolorization	Synergy (CI 0.16–0.28) A	[141]
Lavandula stoechas essential oil + Laurus nobilis essential oil + Mentha pulegium essential oil, 1:1:1, 2:1:1, 1:2:1, 1:1:2	ABTS• decolorization	Synergy (CI 0.25–0.97) A	[141]
Lavandula stoechas essential oil + Mentha pulegium essential oil, 1:2, 1:1.2:1	DPPH• decolorization	Antagonism (CI 1.65–7.04) A	[141]
Laurus nobilis essential oil + Lavandula stoechas essential oil, 1:2, 1:1.2:1	DPPH• decolorization	Synergy (CI 0.21–0.40) A	[141]
Laurus nobilis essential oil + Mentha pulegium essential oil, 1:2, 1:1.2:1	DPPH• decolorization	Synergy (CI 0.24–0.40) A	[141]
Lavandula stoechas essential oil + Laurus nobilis essential oil + Mentha pulegium essential oil, 1:1:1, 2:1:1, 1:2:1, 1:1:2	DPPH• decolorization	Synergy (CI 0.23–0.75) A	[141]
Glutathione + red cabbage extract	FRAP	Synergy T	[78]
Ascorbic acid + red cabbage extract	FRAP	Additivity T	[78]
Trolox + red cabbage extract	FRAP	Additivity T	[78]
Lavandula stoechas essential oil + Mentha pulegium essential oil, 1:2, 1:1.2:1	CUPRAC	Antagonism (1:2, 1:1) A Synergy (2:1) A	[141]
Laurus nobilis essential oil + Lavandula stoechas essential oil, 1:2, 1:1.2:1	CUPRAC	Synergy (CI 0.03–0.19) A	[141]
Laurus nobilis essential oil + Mentha pulegium essential oil, 1:2, 1:1.2:1	CUPRAC	Synergy (CI 0.03–0.27) A	[141]
Lavandula stoechas essential oil + Laurus nobilis essential oil + Mentha pulegium essential oil, 1:1:1, 2:1:1, 1:2:1, 1:1:2	CUPRAC	Antagonism (1:1:1) A Synergy (2:1:1, 1:2:1, 1:1:2) A	[141]
Trolox + red cabbage extract	ORAC	Synergy T	[79]
TEMPOL + red cabbage extract	ORAC	Antagonism T	[79]

^A, ANOVA; ^T, Student’s “t” test.

On the basis of analysis of synergy, Liu et al. conclude that esculetin and gentisic acid may be the principal phenols to enhance the oxidation stability of tea seed oil [70].

8. Cellular Antioxidant Activity

Cellular antioxidant activity, based on estimation of inhibition of oxidation of intracellular fluorescent probe upon treatment with an oxidant of cells preincubated with antioxidants [48,49], is the next step from in vitro cell-free assays towards a situation in the cell. Moreover, they allow study of the effects of antioxidants on other cellular functions, apart from the level of reactive oxygen species. However, they have serious limitations, too. The scavenging of ROS responsible for inhibition of oxidation of fluorescent probes is contributed to by endogenous intracellular antioxidants and antioxidant enzymes. The transport of antioxidants to the cells may affect their intracellular concentrations, which are, in fact, unknown. Exogenous antioxidants may be subject to intracellular metabolism. All these factors lead to differences in studies of the same antioxidants obtained with various cell types.

In rat cardiomyocyte H9c2 cells, the interactions between delphinidin and lycopene, chlorogenic acid and lutein, chlorogenic acid and lycopene, delphinidin and chlorogenic acid, delphinidin and lutein, and lycopene and lutein ranged from synergistic to antagonistic in all cases, depending on the concentration ratio of antioxidants, as demonstrated by cellular activity assay measuring the inhibition of ROS formation [24].

In the cellular antioxidant activity assay using human umbilical vein endothelial cells (HUVECs), Caco-2 human colorectal adenocarcinoma cells, and L-02 cells, the observed effects were cell-line dependent. The interactions between lutein and quercetin were antagonistic, additive, or synergistic in HUVECs and Caco-2 cells and antagonistic or additive in L-02 cells, depending on the concentration ratio of the components. The interactions between lutein and luteolin were synergistic in HUVECs and antagonistic, additive, or synergistic in Caco-2 and L-02 cells. Lycopene and quercetin showed synergy in HUVECs and Caco-2 cells and synergy, additivity, or antagonism in L-02 cells. The interactions between lycopene and luteolin were synergistic in HUVECs, antagonistic or synergistic in Caco-2 cells, and antagonistic, additive, or synergistic in L-02 cells. These differences between cell lines were attributed mainly to the effect of carotenoids on the expression of membrane transporters able to mediate flavonoid uptake by the cells [62].

9. Discussion

In vitro studies are believed to be a necessary initial step for elucidation of the mechanisms of interactions between antioxidants before moving on to in vivo studies. However, the results of these studies should be interpreted with caution. The reaction conditions used in most assays (solvent, pH, temperature, and the absence of a natural matrix and other compounds, such as metal ion chelators) may differ from the physiological conditions of antioxidant action, thereby limiting their physiological relevance (Table 1); e.g., the DPPH^• decolorization assay usually employs methanol or ethanol as the reaction medium [34,35,36], and the FRAP assay is run at a pH of 3.6 [37,38], which is rather far from the physiological pH range. The time frame of the assays is limited, so slow reactions may not be included. Moreover, the oxidants/probes in some antioxidant assays (ABTS^•, DPPH^•, fluorescein in ORAC) are synthetic radicals not occurring in nature, and reduction of these radicals by antioxidants does not necessarily correlate with protection of biologically relevant substrates such as DNA, proteins, or lipids.

In cellular antioxidant activity assays, the concentrations of antioxidants used are usually much higher than those to which cells can be exposed in the organism, and the actual intracellular concentrations of antioxidants during the assay are unknown. One point of criticism is that most studies of antioxidant interactions in vitro are performed using the “black box” approach, not including analysis of reaction products, and not aiming at elucidation of the mechanisms of antioxidant reactions and interactions.

However, the main problem concerns the relevance of these studies to the question of organismal effects of compounds referred to as antioxidants. Endogenous antioxidants and antioxidant vitamins interact with others, but the antioxidant system involves many compounds, much more than those studied in simple in vitro systems. Some of them cooperate with enzymes, being oxidized and regenerated in enzymatic reactions, not included in in vitro experiments.

Other exogenous plant-derived antioxidants (mainly phenolics) are secondary plant metabolites, synthesized for defense against environmental stress [146,147], herbivores [148,149,150] and microorganisms [151,152].

The physiological effects of these compounds are complex. One is the direct antioxidant action. However, most exogenous antioxidants are generally absorbed in minute amounts [153,154,155]. Such amounts of exogenous antioxidants generally cannot significantly affect the redox equilibrium of the human body [156]. Moreover, they are subject to metabolic transformations by gut microbiota and the host detoxification system and elimination. Flavonoid metabolites have different biological and antioxidant properties than their parent compounds, suggesting that data from in vitro studies using nonmetabolized flavonoids are of limited relevance for their in vivo effects [157]. Nevertheless, they can be locally accumulated in higher amounts and affect cellular and tissue redox signaling [158]. NF-κB is a ubiquitous transcription factor that regulates many central events in normal cell function and fate. NF-κB is redox-sensitive, and, in general, oxidants promote its activation while antioxidants inhibit it. Flavanols and procyanidins can interfere with NF-κB activation not only by altering cellular redox state, but also by specific binding to proteins involved in the NF-κB pathway. (−)-Epicatechin and dimers can inhibit NADPH oxidase and the subsequent superoxide production by directly binding to the enzyme [159].

Apart from their chemical reactions with oxidants, antioxidants exert biological effects mediated through signaling pathways and changes in gene expression. (−)-Epicatechin and procyanidins were demonstrated to directly inhibit protein phosphatases, kinases, or other signaling proteins [159]. Pueraria flavonoids induce autophagy via stimulation of the PI3K/Akt/mTOR signaling pathway [160], the PI3K/Akt/HIF-1α signaling pathway to mitigate cancer resistance [161], and the Wnt/β-catenin pathway, important for cancer growth and metastasis [162] and inhibiting the cGAS-STING signaling pathway ameliorating LPS-induced acute lung injury [163], to mention only a few examples.

In biological membranes, flavonoids interfere with the raft structure and thus membrane signaling [164]. Naringenin inhibits acid sphingomyelinase-mediated membrane raft clustering and reduces NADPH oxidase activation in this way [165]. EGCG affects lipid rafts, which blocks the activation of the c-Met receptor and the HGF/c-Met pathway responsible for invasion and metastasis of most human cancers [166].

These effects may be related to or independent of the antioxidant activity of polyphenols. However, interaction with signaling pathways may mediate the antioxidant effects of these compounds at the cellular and organismal levels. There are many reports on the activation of the Keap1-Nrf2 pathway, the primary cellular defense system against oxidative stress, activating genes responsible for the synthesis of antioxidant proteins and mounting endogenous antioxidants (e.g., [167,168,169,170]).

These effects, more important than the direct antioxidant action of phenolics [171,172], escape detection by in vitro assays of antioxidant activity and antioxidant interactions.

10. Conclusions and Perspectives

Taking all above aspects into account, the importance of results of studies of interactions between antioxidants for understanding the in vivo effects of antioxidants seems limited.

However, in vitro studies of antioxidant interactions may be useful for food preservation. Studies of the interactions of antioxidants in the prevention of the oxidation of oils and other fats can be directly translated into practical protocols, allowing for limiting the amounts of antioxidant additives. However, antioxidant interactions characterized in model systems should be validated in the actual food matrix under realistic processing and storage conditions.

Future studies on the in vitro interactions of antioxidants should include thorough analysis of multiple concentrations and ratios, concentration–response dependencies, kinetic analysis, careful selection of predefined additivity models, inclusion of mechanistically distinct assays, and appropriate statistics. They should aim at elucidation of the molecular mechanisms of the interactions, combining antioxidant assays with analytical techniques, enabling identification of intermediates, final products, and byproducts of apparently simple antioxidant assays. They should use substrates and conditions relevant to physiology or food systems. Finally, they require validation in relevant food or biological systems. Understanding the real course and pathways of antioxidant reactions should allow explaining and predicting interactions between antioxidants, also operating in vivo.

Abbreviations

The following abbreviations are used in this manuscript:

AAPH	2,2′-Azobis(2-amidinopropane) dihydrochloride
ABTS	2,2′-Azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)
AMVN	2,2′-Azobis(2,4-dimethylvaleronitrile)
BHT	Butylated hydroxytoluene
CAPE	Caffeic acid phenethyl ester
CI	Combination Index
CTAB	Cetrimonium bromide
DMPC	Dimyristoylphosphatidylcholine
DMPO	5,5-Dimethyl-1-pyrroline N-oxide
DPPH	2,2-Diphenyl-1-picrylhydrazyl
ECG	Epicatechin gallate
EGCG	Epigallocatechin gallate
EPR	Electron paramagnetic resonance
FRAP	Ferric Reducing Antioxidant Power
GSH	Glutathione
HUVECs	Human umbilical vein endothelial cells
ORAC	Oxygen Radical Absorbance Capacity
PMHC	2,2,5,7,8-Pentamethyl-6-hydrochromanol
SDS	Sodium Dodecyl Sulfate
TBHQ	tert–Butyl hydroxyquinone

Data Availability Statement

No new data were created or analyzed in this study. Data sharing is not applicable to this article.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This research was funded by the project “Modification of anthocyanins/anthocyanidins as new markers of food oxidation” (application number 2023/51/B/NZ9/02490) financed by the National Science Centre (NCN), Poland, within the “OPUS 26” program.