Chitosan as an antioxidant alternative to sulphites in oenology: EPR investiga- tion of inhibitory mechanisms

The efficacy against oxidative degradation in model and sulphite-free white wines of two commercial, insoluble chitosans (one being approved for winemaking) were investigated by electron paramagnetic resonance (EPR). Both compounds at various doses significantly inhibited the formation of α-(4-pyridyl-1-oxide)-N-t-butylnitrone (4-POBN)-1-hydroxyethyl adducts under normal wine storage conditions. Pre-incubation with 2 g/L chitosan followed by filtration had a better effect than adding 50 mg/L sulphur dioxide to the experimental Chardonnay wine on the release of 4-POBN adducts after 6 days of incubation with 100 μM iron(II). In a relevant photooxidative system acetaldehyde formation was significantly reduced after 6 days of incubation. Parallel EPR tests were performed to assess the importance of metal chelation (iron and copper) versus direct scavenging of hydroxyl radicals on the effect of chitosan. The present data support the potentiality of using biocompatible chitosan as a healthier complement and/or alternative to sulphur dioxide against white wine oxidative spoilage.


Introduction
Skilled oxidation management is perhaps the most crucial task in winemaking since it can offer significant improvement of the organoleptic characteristics (color, flavors and taste) and shelf life of the finished wine. In the last decades a vast literature has arisen dealing with the molecular events behind non-enzymatic oxidation of wine (Danilewicz, 2003;Oliveira, Ferreira, De Freitas & Silva, 2011;Waterhouse & Laurie, 2006) that would guide technological interventions in wineries. The most plausible scenario for wine oxidation globally resembles that of oxidative stress in biology but, unlike in respiring cells, molecular oxygen is here reduced to water univalently, with successive formation of the hydroperoxyl radical HOO (the protonated form of superoxide (O 2  -) at wine pH), hydrogen peroxide (H 2 O 2 ), and finally the hydroxyl radical (HO). This cascade of reactions are susceptible to occur at any stage of winemaking, including after bottling, since the volume of dissolved oxygen and the headspace above the wine can reach several mL depending on type of closure and adopted vinification technology (Grant-Preece, Barril, Schmidtke & Clark, 2017). This results in oxidation of the wine polyphenols to corresponding o-quinones and brown pigments (a phenomenon called 'browning'), with undesired effects on the aromatic profile and color (Oliveira et al., 2011;Waterhouse & Laurie, 2006). Trace transition metals, particularly iron and copper, have been shown to play a cardinal role in wine oxidation, notably because they catalyze the reduction of H 2 O 2 to HO by a Fenton-type reaction, being then redox cycled by those polyphenols, with experiments often involving the representative compound 4-methylcatechol (4-MeC), not itself found in wine but bearing a typical catechol moiety (Danilewicz, 2003). Finally, HO will oxidise ethanol and tartaric acid to acetaldehyde CH 3 CHO and glyoxylic acid, respectively, the former imparting to white wine a characteristic oxidative odor upon accumulation.

5
Of the available methodologies to study the reactivity of HO in wine oxidation, electron paramagnetic resonance (EPR) spectroscopy coupled to spin-trapping has led to conclusive advances in the understanding of free radical processes. Fig. 1 shows that hydroxyl radicals, which nonspecifically attack any molecule at diffusion controlled rates (i.e., with second-order rate constants > 10 9 M -1 .s -1 ), will oxidise ethanol to the main, thermodynamically stabilized secondary 1-hydroxyethyl radical (1-HER) intermediate. In low O 2 conditions 1-HER is readily oxidized by Fe(III) to yield acetaldehyde. Despite being quenched by many wine constituents such as polyphenols and thiols (Kreitman, Laurie & Elias, 2013) enough 1-HER remains available to be spin trapped on nitrones added to a wine oxidation system, giving nitroxide adducts that can sometimes be detected for days (Elias, Andersen, Skibsted & Waterhouse, 2009a, 2009bElias & Waterhouse, 2010;Kreitman, Cantu, Waterhouse & Elias, 2013;Zhang, Shen, Fan, García Martín, Wang & Song, 2015;Nikolantonaki et al., 2019).
In oenology the most widely used intervention to protect must and wine against oxidation and microbial activity is adding sulphur dioxide (SO 2 ). Bisulfite, the active form of SO 2 at wine pH, is presumed to lessen overall wine oxidation process at several stages, i.e., by scavenging H 2 O 2 to yield sulfate ( Fig. 1), reducing o-quinones back to their phenolic precursors, or binding to carbonyls, especially the most abundant one, acetaldehyde (Danilewicz, 2003;Oliveira et al., 2011). Despite their remarkable efficacy, simple implementation and low cost, however, sulphites have demonstrated latent adverse effects in hypersensitive individuals, e.g., they may aggravate the symptoms of allergic asthma (Vally, Misso & Madan, 2009), and there is movement to promote supposedly healthier non-sulphited wines, or to support antioxidant alternatives in winemaking. Most if not all of the new technologies (pulsed electric fields, ultrasounds, UV irradiation) and natural preservation chemicals (dimethyl dicarbonate, lysozyme, bacteriocins) under development have as their purpose antimicrobial and enzyme inactivating effects (Santos, 6 Nunes, Saraiva & Coimbra, 2012). Furthermore, the use of other additives such as glutathione, ascorbic acid or polyphenols intended at inhibiting free radical mediated wine oxidation at H 2 O 2 stage or downstream remain marginal (Kemp, Alexandre, Robillard & Marchal, 2015).
Another attractive route to control wine oxidation is inactivation of catalytic metals by potent chelators, as such intervention would, in principle, simultaneously inhibit Fenton chemistry, and the formation of o-quinones and acetaldehyde (Fig. 1). Chitosan is a -1.4connected linear polymer of D-glucosamine usually obtained by deacetylation of chitin, a homopolymer of N-acetyl glucosamine extracted from insects, crustaceans or fungi. Due to its regular and high density of amino and hydroxy groups ( Fig. 1), this non-toxic, biodegradable biopolymer has remarkable metal chelation power and, because of its polycationic nature, chitosan also exhibits anti-microbial activity, both properties having attracted food scientists for decades (Bornet & Teissedre, 2008). Moreover, use of chitosan as an additive in winemaking for preventing cloudiness, removal of heavy metals, and reduction of Brettanomyces spp. contamination has been regulated by the EU in 2011 (EC Regulation No 53/2011). Since its early introduction as anti-browning agent in white wines (Spagna, Pifferi, Rangoni, Mattivi, Nicolini & Palmonari 1996) chitosan has stimulated wine researchers as a substitute for SO 2 (Chinnici, Natali & Riponi, 2014) and is gaining popularity in winemaking (Colangelo, Torchio, De Faveri & Lambri, 2018;Filipe-Ribeiro, Cosme & Nunes, 2018;Castro-Marín, Buglia, Riponi & Chinnici, 2018). In the present work established EPR spin trapping and wine oxidation relevant techniques were applied for the first time, to obtain a deeper understanding of the mechanisms of how an approved, insoluble chitosan protects against white wine spoilage in winemaking conditions. 7

2.
Materials and methods

Model wine solution
One litre of model wine solution consisting of 12% (v/v) ethanol and tartaric acid (8 g/L) was prepared and its pH was adjusted to 3.5 with 10 M NaOH. To guarantee air saturation, samples were stirred for 1 h before carrying out the experiment.

Measurement of Fe(II) chelating activity
First, the chelating activity of CHI-2 was determined in model wine (pH 3.5) at room temperature using the ferrozine competition assay (Stookey, 1970) with modifications.
Briefly, 0.1 mL of CHI-2 in suspension at different concentrations (0-10 g/L) was mixed with 50 L of ferrozine solution (0.72 mM) in 1.5 mL Eppendorf tubes. Following stirring of the samples for 10 min in darkness, 0. Second, the unchelated iron content of real wine samples spiked with 100 M Fe(II) alone or in the presence of CHI-2 (0.5 and 2 g/L) was also determined by means of flame atomic absorption according to the relevant OIV method (see above). Briefly, samples were saturated with air and aliquots (20 mL) were placed into 50-mL Falcon tubes sealed with stoppers and continuously agitated for 48 h at 20 °C in darkness. Afterwards, samples were centrifugated (45 g) and filtered prior to injection for iron analysis. The instrument was an Agilent 240FS AA spectrophotometer, with a deuterium lamp for background radiation correction, a hollow cathode lamp at 248.3 nm, and the airacetylene flame.
Calibration curves were plotted using standard iron diluted with deionized water. All analyses were performed in triplicate.

Wine sample preparation
An additional set of iron-spiked wine samples containing CHI-2 (0.5 and 2 g/L) or SO 2 , as potassium metabisulfite, (50 mg/L) were prepared and stored as outlined for flame atomic absorption studies. Samples were placed at 20 °C in a temperature controlled chamber for 16 days at a distance of 5 cm from two cool daylight fluorescent lamps (Sylvania T8 Luxline Plus 36W 840) producing a light at 300580 nm wavelength and an average intensity of 2000 lux. The light intensity was measured using a 51000 series digital lux meter (Yogokawa, Lyon, France). All samples were shaken for 2 min four times/day throughout. All experiments were performed at least in triplicate.

HPLC-DAD analysis of acetaldehyde
At the end of the irradiation period the samples were analyzed for their content in acetaldehyde using a Merck Hitachi HPLC system consisting of an Elite LaChrom L-7000 interface module with a diode array detector (DAD) (L-7455) and a EZchrom workstation for data processing. The UV spectra were recorded in the range 220400 nm. With the aim to detect exclusively the free fraction of aldehydes which take part in oxidation process, no acid hydrolysis of samples was carried out. Samples (800 L) were incubated with 200 L of a DNPH solution (10 mM in 2.5 M HCl) for 1 h at 45º C in darkness. After cooling at room temperature separation of the DNPH adducts was achieved on a Nucleodur C18 Htec column (Macherey-Nagel, Düren, Germany; 250 × 4.6 mm; 5 m) with a flow rate of 0.8 mL/min. Solvent A was acetonitrile; solvent B was water containing 0.05% (v/v) solution of phosphoric acid (pH 2.7). The elution program was the following: 0 min, 40% A, 8 min, 85% A, 9 min, 40% A, 13 min, 40% A, and injection volume was 20 L.
The identification of the observed derivatives was based on their retention time compared with those of standards tested at 360 nm as well as their spectral characteristics.
Quantification was based on peak area.

Oxygen radical absorbance capacity (ORAC) assay
The assay was performed in microplates as previously described (Kandouli et al., 2017

Preparation of solutions and suspensions for EPR spin trapping analysis
Samples to be scanned by EPR were aspirated into 50 L glass capillary tubes (Hirschmann Lab., Eberstadt, Germany), as to fill them completely, and sealed with Critoseal (McCormick Scientific, St Louis, MO) at lower (nucleophilic addition and Fenton reaction with 4-MeC) or both ends (remaining studies).

Solutions for in situ photolysis and Fenton reaction in model wine and calculation of rate constants
Hydrogen peroxide (3% v/v) was used as photolytic precursor of HO. Solutions of chitosan (0.12 g/L) dissolved in water containing 0.5% (v/v) acetic acid (pH 3.18) and DMPO (3.33 mM, final) were continuously illuminated using a 1000 W xenon-mercury UVVis light source (Oriel, Newport Corp., Irvine, CA) guided within the EPR cavity through an optical glass fiber. The corresponding blank spectra were substracted from experimental spectra before data processing.
The apparent second-order rate constant k CHI for the reaction of HO with chitosan was calculated using the equation: where I 0 and I is the intensity of the EPR signal recorded in the control and in presence of chitosan, respectively, C DMPO and C CHI are the concentrations of DMPO and chitosan, respectively, and k DMPO is the second-order rate constant for the trapping of HO on DMPO. The slope of the regression plot of I 0 /I against C CHI for a constant value of C DMPO was used to estimate k CHI (Finkelstein, Rosen & Rauckman, 1980): assuming that k DMPO = 3.4 × 10 9 M -1 .s -1 using the above conditions and photolytic system (Finkelstein et al., 1980).

Suspensions for nucleophilic addition assays
A suspension of tested chitosan (0.52 g/L) in water containing a wine relevant concentration of Fe(III), as FeCl 3 , of 30 mg/L and Cu(II), as CuSO 4 , of 12.5 mg/L was stirred for 1 h at room temperature to allow for metal complexation by chitosan. EPR spectra were recorded 1 min following addition of aqueous DEPMPO (55 mM) to the suspension.

Suspensions for Fenton reaction assays and incubations
All experiments described below (incubations and EPR spectrometry) were conducted at 2022 °C in darkness.

13
To assess the effect of CHI-2 (0.5 or 2 g/L) or SO 2 (50 mg/L) on 4-POBN adduct formation the tested inhibitor was pre-incubated with 100 M Fe(II) for 48 h in model wine.

Statistical analysis
Data are given as mean ± SEM for the indicated number of independent experiments.
Evaluation of statistical significance was conducted by one-way analysis of variance (ANOVA) followed, if significant (p < 0.05), by a posteriori Duncan test. Differences between groups were considered significant when p < 0.05.

Results and discussion
Owing to its insolubility at wine pH and known metal chelation property, suspensions of chitosan were stirred for 2 days in darkness with 100 M (5.5 mg/L) Fe(II), a typical concentration found in white wine, to ensure maximum metal chelation before oxidation under wine conditions was induced. Under these pre-incubation conditions 500 M ferrozine were found to chelate 142 M of Fe(II) in model wine at pH 3.5, while increasing the pH to 4.5 resulted in a 30% increase of the chelating power, in agreement with previous observations (Stookey, 1970).

EPR evidence that chitosan slowers free radical formation during wine oxidation
A general method to follow 1-HER formation in oenology as an intermediate in nonenzymatic wine oxidation (Fig. 1) is to apply EPR spin trapping using the linear arylnitrone 4-POBN as the spin trap (Elias et al., 2009a(Elias et al., , 2009bNikolantonaki et al. 2019 (Halpern, Yu, Barth, Peric & Rosen, 1995;Nakao & Augusto, 1998;Pou et al., 1994). In the same system removal of ethanol allowed transient detection of the 4-POBN/hydroxyl radical adduct (4-POBN-OH) as a sextet with slightly different EPR parameters: a N = 14.99 G, a H = 1.65 G, and g = 2.0057, consistent with early data (Pou et al., 1994). The fact that 1-HER, and not the primarily formed HO (Fig. 1), is the major species trapped in 4-POBN spin trapping studies on alcoholic beverages mainly relies on: (i) the very low stability of 4-POBN-OH versus 4-POBN-1-HER (Halpern et al., 1995;Pou et al., 1994), and (ii) the large excess of ethanol (molar range) with respect to the nitrone (millimolar range) in the system to compete with HO. Indeed, detection of nitrone/HO adducts in oxidizing wine required a molar concentration of the trap (Elias et al., 2009a).
As depicted in Fig. 1 endogenous wine's phenolics can be considered suitable Fe(II)/Fe(III) redox recyclers to sustain the Fenton reaction involved in wine oxidation (Elias et al., 2009a;Elias & Waterhouse, 2010). When the model wine system above had 1 mM of 4-MeC added (taken as a model for wine's catechols), a wine's typical concentration with respect to total phenolics (Kreitman, et al., 2013b), 4-POBN-1-HER adducts developed over 3 days at ambient temperature and in darkness, provided that 16 incubating samples were always well aerated. Given the known high stability and resistance to redox-active agents of 4-POBN-1-HER adducts (Halpern et al., 1995) a sensitive EPR acquisition method was applied here, where accumulating signals in ~7-min blocks allowed detection of weak signal intensities since 1 h in control samples (Fig. 3A).
Both chitosans added in suspension up to 2 g/L significantly inhibited oxidative formation of 4-POBN-1-HER with similar profiles (but no clear dose-dependence), CHI-2 being the most effective after 48 h of action. A very significant inhibitory effect of the best compound, CHI-2, added at 1 or 2 g/L was also seen when incubations were carried out in sulphitefree Chardonnay wine for up to 6 days at ambient temperature, again with no significant dose effect except in the early oxidation phase (Fig. 3B).
In order to address the mechanisms by which chitosans protect synthetic and real wine against free radical mediated ethanol oxidation, i.e., by delaying the formation of 1-HER radical intermediate, incubations in both matrices were carried out in the presence of SO 2 at a winemaking dose (50 mg/L), or the strong iron(II) chelator ferrozine. By interacting with two main components of the Fenton system ( Fig. 1), SO 2 and ferrozine can inhibit 1-HER formation by removing H 2 O 2 or forming iron complexes with no catalytic power, respectively (Elias et al., 2009b;Elias & Waterhouse, 2010;Kreitman et al. 2013b).
The strong decreases in 4-POBN-1-HER formation seen with both types of treatments seem to confirm the pertinence of these two mechanisms (Fig. 3). Thus, spin adduct formation re-increased in SO 2 added samples after 48 h incubation, possibly because decreased levels of free SO 2 (i.e., the scavenging-active SO 2 fraction not linked to acetaldehyde and not already oxidized to sulfate) could no more efficiently eliminate the continuous H 2 O 2 formation in the system.
EPR signals from all samples pre-treated with high ferrozine (500 M) exhibited the lowest intensities throughout the incubation time frame (Fig. 3). This is consistent with the results of the Fe(II) activity assay above suggesting that practically all of the 100 M

iron(II) added should have been complexed by ferrozine into a Fenton-inactive species.
Moreover, from the IC 50 values obtained by the ferrozine assay in model wine it was found that 168 M ferrozine and 2.4 g/L CHI-2 exhibited similar chelating effects toward 100 M of Fe(II). This could explain the similarity of 4-POBN-1-HER inhibition profiles between samples supplied with 150 M ferrozine and those with added CHI-2, but not CHI-1 (Fig.   3A).
The extent to which pre-treating the real wine with CHI-2 had reduced the amount of catalytic Fe(II) before the spin trapping reaction shown in Fig. 3B started, was quantified by flame atomic absorption. The endogenous concentration of Fe in the wine was only of 6.8  0.1 M (both vintages combined). Following addition of 100 M Fe(II) and incubation for 2 days in darkness, 98.9  1.1 M of iron was detected, with a small loss consequent to, e.g., adsorption onto the labware or wine proteins, or chelation by tartaric or citric acids.
In the presence of CHI-2 at 0.5 and 2 g/L, the free iron content of the wine samples was significantly decreased to 48.5  0.4 and 31.2  0.8 M, respectively.
It is therefore possible that part of the effect of chitosans found in the above EPR experiments may be due to Fe(II) chelation properties. Chelation capacity of chitosan in oenology has already been reported (Bornet & Teissedre, 2008;Chinnici et al., 2014;Colangelo et al., 2018). Since for these compounds metal removal is based on the formation of a complex involving amine or hydroxyl groups (Fig. 1), chelation capacity increases with increasing degree of deacetylation and decreasing molecular weight as a consequence of greater availability of amino groups toward metal ions (Bornet & Teissedre, 2008). These structural features may explain the lower effectiveness of CHI-1 versus CHI-2 (Fig. 3A). Furthermore, chitosan can adsorb polyphenols into its matrix, decreasing their level in wine (Chinnici et al., 2014;Spagna et al., 1996). Hence, such a decrease in the 4-MeC (model wine) or oxidizable polyphenols (real wine) contents would indirectly inhibit spin adduct formation by lowering H 2 O 2 levels.

Quantification of H 2 O 2 levels occurring during wine oxidation
In the above experiments the relative EPR intensities were found to be similar over the incubation time frame regardless experiments were performed in 4-MeC-supplemented model or real wine (Fig. 3). To estimate the H 2 O 2 concentrations implicated, a Fenton assay was run in unsupplemented model wine by measuring 4-POBN-1-HER levels, obtained using an identical temperature and acquisition protocol, as a function of Fe (II) and H 2 O 2 constituents. As seen in Fig. 4, spin adduct concentration, which increased with any of these two constituents, was more dramatically affected upon doubling iron(II) content than when H 2 O 2 concentration was increased 10 times. This substantiates the above and previous findings (Bornet & Teissedre, 2008;Elias et al., 2009b;Kreitman et al., 2013b) that decreasing metal ion content in wine may be a more sustainable strategy against oxidation than temporarily scavenging H 2 O 2 by adding SO 2 . Furthermore, in completely filled and stopped capillaries, 4-POBN-1-HER intensities only moderately augmented 10 min versus ~2 min after triggering the Fenton reaction, and consequently the generation/detection system run here can be considered as a controlled one.
As seen in Fig. 3, accumulation of long-lived 1-POBN-1-HER adducts resulted in average EPR intensities peaking at ~2.5 relative units in the controls when 100 M Fe(II) was used to start oxidation. According to the results of Fig. 4 where H 2 O 2 was added at once in the system, this suggests that the total H 2 O 2 concentration decomposed by the Fenton reaction over 36 days under wine oxidation conditions was very low, ranging 0.252.5 g/mL (775 M), as visualized by the dashed line in Fig. 4. Fenton generators commonly used in wine oxidation spin trapping studies involve similar Fe(II) concentrations but at least fourfold higher H 2 O 2 concentrations (Elias et al., 2009a;Nikolantonaki et al., 2019). Obviously, the EPR spin trapping technique applied here underestimates H 2 O 2 levels produced in wine oxidation because a variety of scavenging mechanisms are operating, e.g., reactions with SO 2 . Thus in a set of red wines oxidized in air at 40°C using 100 M Fe(II), a rate of H 2 O 2 formation of ~14 M/30 min was reported (Héritier, Bach, Schönberger, Gaillard, Ducruet & Segura, 2016).

Effect on Fenton-derived 1-HER
Having defined the combination of 100 M Fe(II) + 2.5 g/mL H 2 O 2 (74 M) as a wine- like Fenton system to model incubations of Fig. 3A, it was applied in model wine  4-MeC (1 mM), alone or in the presence of inhibitors, and the effects on 4-POBN-1-HER formation were monitored for up to 1 h. In unsupplemented medium EPR signals, which expectedly increased along with the continuous formation of HO radicals, showed a 2.53.5 amplification in the presence of 4-MeC (Fig. 5A). A similar effect has been reported by Elias and Waterhouse (2010) who suggested that the recycling of Fe(III) to Fe(II) by 4-MeC may increase adventitiously the amount of 1-HER available for spin trapping (Fig. 1).
The control EPR signals in Fig. 5A were decreased up to 89% (unsupplemented) or 95% (with 4-MeC) in samples pre-incubated with CHI-2 suspensions (0.5 or 2 g/L) for 2 days, with no clear dose-response effect. In these experiments background 4-POBN-1-HER adducts were detected in SO 2 added samples up to 30 min.

20
Having investigated the iron(II) chelating property of chitosans as a key step of the inhibition of HO-mediated wine oxidation, additional spin trapping experiments were carried out in attempt to delineate the specific HO scavenging behaviour of these compounds. Previous EPR investigations of the antioxidant properties of chitosan have generally focused on hydrosoluble derivatives and standard assays, including Fenton reaction-based tests on HO scavenging (see, e.g., (Park, Je & Kim, 2003)) for which there is clear interference with Fe(II) chelation property. Here, an iron independent method for producing HO spin adducts was used, where in situ photolysis of 3% H 2 O 2 in the presence of the cyclic nitrone DMPO (ca. 3 mM) in 0.5% acetic acid solution (pH 3.2) afforded the known DMPO/hydroxyl radical adduct (DMPO-OH), giving a characteristic 1:2:2:1 EPR quartet with a N = a H = 14.95 G and g = 2.0053 (Fig. 2b).
Both chitosans, soluble in the medium up to 2 g/L, dose-dependently inhibited the formation of DMPO-OH. Using the kinetic analysis of (Finkelstein et al., 1980) plots of I 0 /I against concentration were obtained (see Methods), exhibiting satisfactory linear fits (Fig.   5B). Assuming an average molecular weight of 120 and 20 kDa for CHI-1 and CHI-2, respectively, second-order rate constants for the reaction of HO were calculated as 7 × 10 12 and 10 12 M -1 .s -1 for CHI-1 and CHI-2, respectively. Such high values, reflecting diffusion-controlled processes, have been reported for many macromolecules, including proteins (Bailey et al., 2014). Using pulse radiolysis, a technique more specific for determining HO rate constants, a value of 6.3 × 10 8 M -1 .s -1 has been reported for deacetylated chitosan from a crustacean, krill (Euphausia superba) at pH 3 (Ulanski & von Sonntag, 2000).
As displayed in Fig. 1, molecular mechanisms for HO scavenging by chitosan backbone can involve either free amine groups and/or their ammonium derivatives, or typical H-abstraction reactions along the polysaccharide unit (Xie, Xu & Liu, 2001).
Moreover, earlier EPR and pulse radiolysis studies have revealed a low selectivity for H-abstraction onto the chitosan unit, i.e., these compounds would behave as if a single preferred site was submitted to HO attack (Ulanski & von Sonntag, 2000). This could explain the linear variations of Fig. 5B, with intercepts of 1.1 and 1.5 for CHI-1 and CHI-2, respectively, close to the theoretical value of 1 in the kinetic model of (Finkelstein et al., 1980).
In another approach to discriminate between iron chelation and free radical scavenging in the inhibitions seen in Fig. 3 Stockham et al., 2013), is yet twice as low as that for 4-MeC containing synthetic wine (Table 1) (Table 1). Therefore, further decrease of this ratio found up to day 3 may suggest a shift from dominant radical scavenging to other inhibitory mechanisms, e.g., delayed Fe(II) chelation by 4-MeC or other wine phenolics. In samples incubated with CHI-2 (2 g/L) the same AUC analysis, yielding expected lower values, also demonstrated a nearly constant 2:1 ratio. This suggests that, once iron had been removed by the 2-days pre-treatment, it is the antioxidant property of chitosan that could have caused the lower spin adduct levels observed. It is worth mentioning, however, that good correlations between ORAC assay and EPR spin trapping have been only reported for peroxyl, but not hydroxyl radicals (Kameya, Watanabe, Takano-Ishikawa & Todoriki, 2014).

Effect on cupric and ferric ions
Adding copper, as Cu(II), during the winemaking process is a common practice, in particular to decrease the levels of sulfur containing compounds responsible for off-flavors before bottling of white wines. Since cupric ions can catalyze H 2 O 2 degradation by a Fenton-like mechanism (Hanna & Mason, 1992), they would potentiate the effect of Fe(III) in wine oxidation (Danilewicz, 2003). In wine studies, however, use of spin trapping is complicated because Cu(II) often induce degradation of nitrones into unwanted nitroxides and/or lead to paradoxical 1-HER formation profiles (Elias et al., 2009b).
In this study EPR has been used to assess indirectly the effect of chitosans on iron and copper at wine-like concentrations, by measuring the impact on nucleophilic induced spin adduct formation. Thus, by forming a transient complex at the nitronyl oxygen of spin traps such as DMPO or its phosphorylated analog, DEPMPO, Fe(III) or Cu(II) catalyze the nucleophilic addition of water to form the corresponding hydroxyl radical adduct, and this reaction is inhibited by Fe(III) (Culcasi et al., 2006a) and Cu(II) (Hanna & Mason, 1992) chelators. When an aqueous solution of DEPMPO (55 mM), a chiral molecule (Fig. 2), was added to a mixture of wine-like 30 mg/L Fe(III) and 12.5 mg/L Cu(II), a major 8-lines EPR spectrum was observed (Fig. 2c) accounting for 24% of the total signal. The cis:trans ratio in Fig. 2c was of 36:64, consistent with a HO scavenging-unrelated, nucleophilic addition mechanism (Culcasi et al., 2006b). When metal added solutions were stirred for 1 h at ambient temperature in the presence of varying amounts of chitosan, further addition of DEPMPO to the suspensions led to dose dependently decreased DEPMPO-OH levels, by 7392%, with no differences between chitosans (Fig. 5C). Altogether these results demonstrated sequestration of Fe(III) / Cu(II) as another facet of the inhibitory action of chitosan in wine oxidation.

Effect on photooxidation-induced acetaldehyde formation
To substantiate the effect of CHI-2 seen in Fig. 3B the production of acetaldehyde was monitored by HPLC-DAD in experimental wine spiked with 100 M Fe(II) and irradiated with fluorescent light (300580 nm) for up to 6 days at ambient temperature. Long term exposure to sunlight or fluorescent tubes has been shown to contribute to the development of browning and the formation of off-odors in white wine. In wine conditions (pH and metals) a main proportion of carboxylic acids in wine, such as tartaric and lactic acids, exist as Fe(III) carboxylate complexes, the irradiation of which leads to a range of carbonyls, including acetaldehyde. This in turn will release free Fe(II), providing an additional source of catalytic iron to fuel the Fenton system, a forced oxidation mechanism termed as 'photo-Fenton' (Grant-Preece et al., 2017).
At opening, acetaldehyde concentration of experimental SO 2 free wine was 7.6  0.7 mg/L (mean value from both vintages), falling within the lowest acetaldehyde concentrations reported in just finished, sulphited dry white wines (Jackowetz & de Orduña, 2013). Following initial production by yeasts during fermentation acetaldehyde can be further synthesized from ethanol through Fenton oxidative degradation (Fig. 1).
Here this second source of acetaldehyde is likely poorly active since when the wine was stored for 2 days in darkness with a 1.5 times air volume in the headspace, a non significant increase to 9.1  0.2 mg/L was found, possibly because catalytic iron present in the wine was only ~0.4 mg/L. Wine samples added 5.5 mg/L Fe(II) and incubated in darkness for 2 days, which retained 5.4 mg/L iron after filtration, showed, however, non significantly increased acetaldehyde levels of 10.8  0.4 mg/L (Table 1). Under photo-Fenton conditions acetaldehyde in the controls increased significantly afterwards, doubling after 6 days irradiation. In wine samples spiked with 5.5 mg/L Fe(II) and having had their iron content lowered by 51% and 68% after 2 days in contact with CHI-2 at 0.5 and 2 g/L, respectively, this irradiation-induced elevation of acetaldehyde concentration was significantly inhibited, with decreases of 19% and 38%, respectively, at day 6 ( Table 1).
Being a strong binder for sulphur dioxide (Oliveira et al., 2011) free acetaldehyde expectedly exhibited the lowest concentrations in irradiated wine added SO 2 (50 mg/L).
However, after 6 days of light exposure, once complete oxidation and/or binding of SO 2 was reached, acetaldehyde production in those samples was not statistically different from that in samples containing 2 g/L CHI-2 (Table 1) and therefore the acetaldehyde inhibition pattern in CHI-2 added wine paralleled that seen for 1-HER formation in Fig. 3B.

Conclusion
The results of this study strengthen current interest in using chitosan as a substitute for and/or complement to lower sulphur dioxide and suphites in winemaking. By monitoring the formation of spin trapped 1-HER, a pivotal intermediate of wine oxidation, EPR analysis sought to establish a chronology of chitosan antioxidant action under wine relevant doses, application and aging conditions. It was found that once the catalytic activity of the metal pool in wine, especially Fe(II)Fe(III), has been partly deactivated by chelation, direct scavenging of oxidizing species such as HO continuing to form at slow rates may represent a significant inhibitory mechanism of chitosan. In this regard, the well documented metal ions-sensitive depolymerization of chitosan by H 2 O 2 (Chang, Tai & Cheng, 2001) could be an additional protective effect against wine oxidation as depicted in Fig. 1. Studies are in progress to verify, using specific tests, if a related free radicalindependent mechanism could participate in the effects seen in the present study.
Of note, the significant impact of chitosans against free radical formation seen here was obtained as the compounds were directly added in suspension in the finished wine.
This will encourage designing future spin trapping studies using EPR techniques specific for large heterogenous samples to follow in situ oxidation of white musts during the alcoholic fermentation.