You are on page 1of 7

Analytica Chimica Acta 589 (2007) 5965

On-line selective detection of antioxidants free-radical scavenging activity based on Co(II)/EDTA-induced luminol chemiluminescence by ow injection analysis
Dimosthenis L. Giokas, Athanasios G. Vlessidis , Nicholaos P. Evmiridis
Laboratory of Analytical Chemistry, Department of Chemistry, University of Ioannina, 451 10 Ioannina, Greece Received 7 November 2006; received in revised form 6 February 2007; accepted 15 February 2007 Available online 22 February 2007

Abstract This study establishes a new method to analyze the radical scavenging activity of antioxidants based on the luminol-H2 O2 -Co(II)/EDTA chemiluminescence and ow injection analysis. The method is based on the catalytic oxidation of hydrogen peroxide by Co(II)/EDTA complex, forming a free radical ux that can produce a stable chemiluminescence signal which is attenuated in the presence of antioxidants. A properly designed FIA manifold and the appropriate regulation of the chemiluminescence-reagent mixture enabled the establishment of a reaction-sensitive analytical procedure that minimizes oxidant-antioxidant interactions while favors the inhibition effect of antioxidants on the free radicals ux. In that manner, the uncontrolled experimental variability induced by side-reactions occurring antagonistically is reduced. The method was examined in-vitro for the continuous monitoring of the generation of oxygen-derived free radicals and antioxidants, which is closer to in-vivo conditions, with three common antioxidants (ascorbic acid, glutathione and uric acid). All three antioxidants were found to inhibit the luminescent signal with strict logarithmic linear mode, yielding calibration curves rectilinear in the range of 5 108 5 105 mol L1 and detection limits at the 108 mol L1 levels. The F-statistic was employed to assess the ability of the method to detect differences in the activity of the examined antioxidants. The results suggest that the proposed method can be used efciently for the detection of free radical activity in real samples. 2007 Elsevier B.V. All rights reserved.
Keywords: Cobalt; Chemiluminescence; Antioxidant; Flow injection; Free radical

1. Introduction Recent advances in the research of free radical and ROS (reactive oxygen species) extend our knowledge on the function of these high reactive chemical species on human body [1]. Oxygen-derived free radicals such as superoxide anion radical (O2 ) and hydroxyl radical ( OH) are thought to be linked to the onset of various pathological conditions such as coronary heart disease, rheumatoid arthritis and cancer [2]. Free radicals and ROS are controlled in biological systems by some enzymes possessing antioxidant activities such as superoxide dismutase and peroxidase [3]. It is well known that non-enzymatic substances also scavenge free radicals and play an important role on protecting cells, tissues and organs from harmful effects of oxygen-derived free radicals and ROS [4].

Corresponding author. Tel.: +30 2651098401; fax: +30 2651044831. E-mail address: avlessid@cc.uoi.gr (A.G. Vlessidis).

The non-enzymatic antioxidants include powerful antioxidants such as uric acid, glutathione and ascorbic acid, some moderately reducing compounds like glucose, which occasionally attend the redox reactions in body, and some reductive drugs [5]. Assessing the scavenging activity of antioxidants to free radicals or ROS is important to assess the pathological state, evaluate the damage made by free radical and ROS, and provide evidence for the generation of free radical and ROS in the human body [6]. For this reason several methods have been presented dealing with the determination of the scavenging activity of antioxidants. Of these methods, chemiluminescence (CL) has attracted the attention of many researchers for the determination of free radical and ROS because of its sensitivity, convenience and simplicity [79]. Up to date, many reports have been published for the assessment of the scavenging activity of antioxidants with the CL method. In most occasions cited inside the literature luminol is used to produce a strong CL signal, which is then scavenged by antioxidants, the difference between initial and nal signal

0003-2670/$ see front matter 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2007.02.041

60

D.L. Giokas et al. / Analytica Chimica Acta 589 (2007) 5965

being a measure of antioxidant activity [819]. However, a certain degree of variability stems from the fact that the measured inhibition of chemiluminescence is partly due to the consumption of free radicals by the antioxidant and partly due to the reaction between the antioxidant and the oxidant. Thus, the measured attenuation of signal reects the rates of both reactions and the reported results strongly depend on the ability of the system to account for both reaction rates. In most occasions, the analytical procedure does not allow controlling these reactions experimentally, thus the measurements reect the assumption that the variation of the response of the depended variable (concentration of the antioxidant) is the same in both CL-inhibition reaction patterns. In the absence of antioxidants this assumption is valid since only one mechanism is responsible for the observed signal. In the presence of antioxidants though there is a certain degree of variability stemming from extraneous variables (scavenging ability of each antioxidant, reaction with oxidant, etc.) that cannot be accounted for by the experimental protocol especially when mixtures of antioxidants are present in the same sample, as usually occurs in real samples. Another source of uncertainty stems from the fact that not all antioxidative compounds quench radicals equally well, while some of them (individually or in mixtures) do not show signicant antioxidant protective activity against OH or O2 species which are generally considered to be the ROS of principal physiological relevance in tissue damage in vivo [12]. As a result samples presenting the same free radical scavenging activity can differ notably in their total antioxidant reactivity due mainly to the differences among antioxidants to donate a hydrogen atom in the reaction with a free radical thus scavenge more or less effectively the available free radicals. For this reason various methods may exhibit quantitatively different values of antioxidant activity for the same sample [12]. This is also reected in the strong scattering of the data between total radical-trapping potential (TRAP) and total antioxidant reactivity (TAR) as a function of sample composition [11]. With the above limitations in mind, this study presents a chemiluminescence reaction system that enables the selective determination of free-radical scavenging activity of antioxidant compounds. The method is based on the property of transition metal ions to catalyze the oxidation of hydrogen peroxide towards the formation of reactive oxygen species (mostly OH and O2 which are of physiological and pathological signicance) via the superoxide FENTON reaction. The luminol/H2 O2 /Co(II)/EDTA-chemiluminescence system was decided over the previously established Fe-based system because it offers better signal response with reference to time stability, light intensity, signal reproducibility and correlation coefcients [18]. The reaction was transferred to an on-line (FIA) manifold appropriately designed to exploit the inhibition effect of antioxidants on the free radicals while minimizing the effect of side-reactions generated from the oxidant-antioxidant interaction. The ability of the method to detect differences in the radical scavenging activity of the target compounds was determined by treating the experimental data of the produced calibration curves with the F-statistic. The results suggest that the proposed method affords a satisfactory assessment of free-

radical scavenging activity of antioxidants species in-vitro, yielding detection limits at the nmole L1 levels and providing a considerable improvement in the reproducibility. 2. Methods and reagents 2.1. Reagents and stock solutions Sodium hydroxide (NaOH), concentrated hydrochloric acid (10 M), boric acid, H2 O2 30% (w/v) and ascorbic acid were obtained from Merck (Darmstadt, Germany). Luminol, uric acid and glutathione (reductive form) were purchased from Sigma-Aldrich (Greece). Cobalt(II) sulphate and ethylenediaminetetraacetic acid disodium salt dihydrate (EDTA) were purchased from Fluka Chemie AG (Switzerland). Buffer solutions, luminol-Co(II)/EDTA (CL-reagent) solution and H2 O2 solution, were prepared weekly in doubly distilled water and stored at 4 C. Prior to use they were allowed to reach room temperature. Fresh antioxidant solutions were prepared daily in doubly distilled water. 2.2. Apparatus A ow injection manifold was used with inlet tubes for feeding the CL-reagent, the oxidant and the sample carrier solutions respectively. The manifold consisted of a peristaltic pump (Gilson, Minipuls 3) with a head accommodating four pump tubes and was implemented with a mixing coil with glass beads for the mixing of the CL-reagent with the oxidant prior to the merging point of the sample solution. The merging point of the CL-light emitting solution with the sample solution is just before the snail-shell like cell-inlet. The detection system consists of the cell, the photomultiplier tube (THORN EMI 9865 B) and the recorder (Linear 1200). The power supply for the photomultiplier tube to detect CL was set at 1.4 kV. The cell is positioned in front of and close to the window of the photomultiplier tube. A schematic diagram of the experimental set-up is shown in Fig. 1. 2.3. Design of FIA Manifold The modied Sariahmetoglu et al.s FIA manifold [19] was initially used. According to this conguration, the injected antioxidant sample is mixed with the oxidant solution before it enters the ow cell and the combined solution merges with

Fig. 1. Experimental set-up of the FI/CL-detection apparatus. Abbreviations: L: luminol-Co(II)/EDTA solution. H: H2 O2 solution. C: carrier solution (0.05 M sodium borate buffer, pH 9). M: mixing tube. F: ow Cell. S: manual sample injector (100 L) in six-port valve. PMT: photomultiplier tube, P: peristaltic pump. RD: recorder. W: waste.

D.L. Giokas et al. / Analytica Chimica Acta 589 (2007) 5965

61

luminol just in front of the ow cell. This design exploits the consumption of oxidant by the injected antioxidants, which results in lower CL signal intensity compared to that observed in the absence of antioxidants. However, in our system, this set-up dramatically impaired the sensitivity of the analysis. This discrepancy is attributed to the chemiluminescence mechanism involved in the proposed reaction pattern (Appendix A). Transition metal ions can drastically increase the CL-signal of luminolH2 O2 system due to the oxidation of the metal by H2 O2 forming a highly reactive and transient free radical product via the superoxide-driven FENTON reaction [18,20]. As a result of these reactions, the CL-signal reaches its peak value in extremely short time and rapidly decays. In the presence of a metal chelator though there is an equilibrium regime between the metal and the chelator, which serves as a constant source of free radicals to the system [20,21]. As a result, a prolonged light emission is produced. In the Sariahmetoglu et al.s, (2003) [19] manifold conguration, mixing of the oxidant with the Co(II)/EDTA/luminol solution immediately before the ow cell resulted in a very high CL signal which was insensitive to the presence of antioxidants unless a very high concentration has been injected. When Co(II)/EDTA was omitted from the reaction mixture the sensitivity was improved but the dilution of the oxidant mixture did not enable a considerable reduction in the detection limits. Increasing H2 O2 (in order to compensate for the dilution) did not provide any improvement possibly due to uncontrolled reaction rate between oxidant-Co(II)/EDTA and antioxidantoxidant solution, which gives an extra error in the difference between baseline and inhibition signal. To cope with these problems, a different conguration was examined. According to the new design the sample-carrier solution is pumped into the merging point with the mixed oxidant/luminol/Co(II)/EDTA solution at the inlet of the ow cell, while oxidant solution mixes with the luminol-Co(II)/EDTA solution previously. This set-up is based on the notion that the equilibrium between Co(II)/EDTA, acts as a Co(II) ion pool by releasing Co(II) ion at a relatively stable rate (Appendix A). This process guaranties the formation of a hydroxyl free radical ux in a reasonably constant rate and therefore feeds the CL-system with a constant concentration of free radicals for a relatively prolonged time span. As a result of this reaction pattern, a constant CL signal is produced which is stable for about 30 s [18,21]; a time-period more than adequate to deliver the photoemissive luminol oxidation mixture to the photomultiplier cell. Because of the very high reactivity of free radicals, their reaction with the antioxidants is very rapid thus overwhelming the reaction between antioxidant and oxidant. In that way, the interference of oxidant is ignored and the signal only reects the scavenging activity of antioxidants to the free radicals. Furthermore, by appropriately regulating the H2 O2 concentration and the ow rate the blank signal is diminished thus improving the sensitivity of the analysis. 2.4. Procedure On switching on the pumps, the CL-reagent stream (L in Fig. 1) and the carrier solution (C in Fig. 1) are initially mixed in

order to reach a stable background. Subsequently, line H (Fig. 1) merges the oxidant (H2 O2 ) with the CL-reagent stream immediately before a mixing coil (M in Fig. 1) and the combined ow is merged with the buffer reagent in order to obtain the baseline signal. The total ow rate is 1.5 mL min1 , shared equally among the three channels. Upon injection of the sample, a loss of signal (negative peak) is observed which corresponds to the CL scavenging activity of the sample (antioxidant). Thus, the effects of antioxidants were measured by the depression of the signal from its initial (uninhibited) level and expressed as percentage attenuation of the CLmax , according to Eq. (1): (%) Inhibition = peak height 100% (baseline background) (1)

Each sample was injected three times on time intervals of 30 s and the nal result was the mean of triplicate measurements. 3. Results and discussion 3.1. Optimization of the working conditions It is well known that the characteristics of the CL signal are pH-dependent and are sensitive to its changes. However, in the proposed reaction scheme pH is also critical for the complexation of Co(II) with EDTA. Therefore, pH was the rst parameter evaluated for its effect on the analytical response. Boric acid (0.05 M)/NaOH was used for regulating the pH of the solutions due to its large buffering capability, compared to some other buffering systems [18,21]. In concurrence with previous studies using the same CL reagent mixture, less alkaline reaction conditions (pH 99.5) favor the production of an intense and stable CL signal (Fig. 2). Although at higher pH values the signal is increased, the baseline also increases, which decreases the sensitivity of the analysis. Furthermore, the reproducibility suffers which is not surprising because too alkaline conditions favor the decomposition of H2 O2 leading to the formation of bubbles in the FIA tubing [22].

Fig. 2. Effect of pH on the luminescence signal.

62 Table 1 Optimal analytical parameters Analysis mode Batch [18] On-line (this study) Buffer Sodium borate (0.05 M) Sodium borate (0.05 M)

D.L. Giokas et al. / Analytica Chimica Acta 589 (2007) 5965

pH 9.009.50 9.00

Flow rate Not applicable 1.5 mL min1

Luminol (mg L1 ) 100 20

H2 O2 (M) 5 103 1 104

Co(II) (mg L1 ) 2 100

EDTA (mg L1 ) 10 300

At pH values below 8.5, the signal is signicantly attenuated not only because of the weak oxidation strength of the oxidant but also due to the insufcient complexation of Co(II) with EDTA [23]. A buffer solution with pH 9 was therefore considered to be optimal and chosen for further work. Although the variables dening the CL process can be assumed to be independent within a narrow range of concentrations [24], the modication of a method from batch to on-line conditions may bring about signicant changes due to alteration of the kinetics and the yield of the reactions. On this basis, the concentrations of the CL reagents were optimized targeting on the optimum analytical signal that does not produce neither too low nor too high a CL signal which can both adversely affect the sensitivity of the CL detector [24,25]. The optimum values are gathered in Table 1, along with the optimum values for batch application. As can be seen, the optimum concentrations for both luminol and H2 O2 were lower than those required for batch analysis. This behavior is probably a result of the lower sensitivity of the batch luminometer as compared to ow-through luminometer. On the other hand, the concentrations of Co(II) and EDTA were signicantly higher but according to the mechanism of the CL reaction (Appendix A) their concentrations strongly depend upon the availability of oxidant (H2 O2 ). In batch conditions, low oxidant concentrations are necessary in order to decrease the rate of its consumption and ensure a low rate of radical production, which is necessary for the evaluation of antioxidants at micromolar and submicromolar concentration levels [11]. In ow conditions, the presence of low oxidant (H2 O2 ) concentrations and the abundance of Co(II)/EDTA ensures the consumption of the oxidant and the production of a constant and intense free radical ux that is rapidly scavenged by antioxidants without residues of oxidant that could interfere with the radicalantioxidant reaction. This mechanism was veried by increasingly surpassing the optimum H2 O2 concentration at the optimum concentrations of the other reagents (luminol = 20 mg L1 , Co(II) = 100 mg L1 and EDTA = 300 mg L1 ). Evidently, the signal became unstable and the reproducibility was deteriorated which is ascribed to uncontrolled reaction rate between oxidant and Co(II)/EDTA as well as to the interference of the excess of oxidant on the radicals-antioxidants reaction. Beyond CL reagents concentration, ow rate denes to a great extend the sensitivity of the analysis because it determines the time available for the detector to monitor the CL-emission when the mixed solution reaches the PMT cell. Thus, different ow rates produce different signals as a function of the monitoring times of the CL-emission. The results depicted in Fig. 3 suggest that the proposed detection system is free from operation factors, such as ow rate. A two-fold change in the ow rate, from 0.5 to 0.9 mL min1 ,

results in a baseline-signal change of less than 10% (Fig. 3) although higher ow rates should normally abridge the residence time of the sample in the PMT cell, thus reduce the analytical response. The relatively stable signal observed with increasing ow rates (at part of the curve in Fig. 3) indicates a time-independent reaction, which is in agreement with previous studies suggesting that the CL-emission from this system remains constant for a reasonable period of time [18]. For lower ow rates the signal is reasonably decreased possibly due to reduction in the reaction rates. Based on these ndings, a ow rate of 0.5 mL min1 for each channel (total 1.5 mL min1 ) was nally selected in order to ensure sensitivity and increase sample through-put. 3.2. Effect of individual antioxidants on CL-signal According Fig. 4A, the three antioxidants displayed strong scavenging activity to free radicals exhibiting a nearly perfect logarithmic relationship between CL-signal loss and sample concentration. The linear range of the calibration curves is extending up to an order to magnitude (Table 2) which is an outcome of the synergistic action of two factors. The selective and rapid reaction of antioxidants with free radicals ( OH, L and O2 ) and the manifold conguration that diminishes the interference due to other reactions between antioxidant and oxidant (H2 O2 ) or peroxides (LOOL , LOO ). The advantages of these reaction patterns are reected in the analytical features gathered in Table 2. In comparison to previous studies, were similar linear logarithmic relationships between CL-signal inhibition and concentration were observed [18,19], linearity is improved and the detection limits are brought down to the

Fig. 3. Effect of ow rate on the analytical response.

D.L. Giokas et al. / Analytica Chimica Acta 589 (2007) 5965

63

Fig. 4. Graphical representation of least squares linear t of signal inhibition to antioxidant concentration. (A) Comparison of the three regression lines. (B) Regression line of uric acid. (C) Regression line of ascorbic acid. (D) Regression line of glutathione. The range inside the dashed lines indicates the 95% condence interval of the three regression lines, respectively. Table 2 Summary of the analytical features of the method Antioxidants Uric acid Ascorbic acid Glutathione Detection limit (nM) 10 10 10 Correlation coefcient 0.9986 0.9989 0.9890 Logarithmic linear equation LI = 15.14ln C + 54.96 LI = 14.17ln C + 54.47 LI = 14.47ln C + 60.58 Linear range 50 nM50 M 50 nM50 M 50 nM50 M IC50 (M) 0.72 0.12 0.73 0.12 0.49 0.22 Number of measurements (n) 5 5 5

LI: chemiluminescence inhibition (%). C: concentration of antioxidants (M). IC50 : Predicted (from the regression curves) concentration of antioxidants needed to decrease the initial chemiluminescence signal by 50%.

nmole L1 levels, thus offering a considerable improvement. The precision of the method was checked by means of repeated assays (n = 5), whose coefcients of variation were calculated (relative standard deviation) to be less than 3.1%, which is very satisfactory. An additional advantage is the high reproducibility which was below 2% as evidenced by sequential injections of the same sample (n = 7) at time intervals of 30 s. 3.3. Comparison of scavenging activity of antioxidants In order to compare the free-radical scavenging activity of the examined antioxidants, the F-test examination was deployed to infer whether there are signicant differences among the slopes of the regression lines obtained for the three antioxidants. The results from this statistic treatment show that the regression lines of these three antioxidants have identical slope values (Fig. 4 and Table 3) in spite of their quite different structures and distinct reaction-mechanisms with free radicals. A possible explanation of this phenomenon is that they all are powerful

antioxidants and react with free radicals very rapidly and quantitatively. The fast reaction rate masks most kinetic differences among these three antioxidants that could be reected in the slopes of the regression curves. Although the kinetics of the radical scavenging reactions may be the same, the amount of free radicals quenched by

Table 3 F-test results of the regression curves Item Slope F Total (U*A*G) U*A U*G A*G 0.36 2.18 0.24 0.05 Prob(>F) 0.7078 0.1779 0.6392 0.8303 H0 Accept Accept Accept Accept Intercept F 5.17 0.01 5.48 6.1 Prob(>F) 0.0208 0.9056 0.0439 0.0356 H0 Reject Accept Reject Reject

Slope: the slope of regression line. Intercept: the intercept of regression line. U: uric acid. A: ascorbic acid. G: Glutathione. H0 is the test hypothesis assuming no signicant difference among the items compared.

64

D.L. Giokas et al. / Analytica Chimica Acta 589 (2007) 5965

Fig. 5. The photoemissive mechanism of luminol oxidation by H2 O2 /Co(II)/EDTA and antioxidant inhibition pathways. LH : monodissociate luminol, L : diazamiquinone radical, LH2 : 5-aminohyphenthalazine- 1,4dione, LOO : hydroperoxide anion, LOOL : endoperoxide, AP: aminophthalate.

each antioxidant depends on the stoichiometric equivalence of antioxidant-to-scavenged radical. It has been estimated that ascorbate and urate have values of 1.7 and 1.3, respectively with peroxyl radical [16]. This means that one molecule of ascorbate and urate can scavenge 1.7 and 1.3 peroxyl radicals, respectively. Although the very rapid reaction rate can accommodate these differences and suppress any kinetic dissimilarities, within the linear range of the regression curve, the different stoichiometric ratio will inevitably be reected in the nal signal since the amount of scavenged radicals inevitably differs. The regression plots of Fig. 4 clearly show this tendency and especially for glutathione whose regression line is slightly shifted above the lines of uric and ascorbic acid. The F-test results show that this behavior is reected in the intercepts of the regression curves. As can be seen, the intercepts of the regression lines do not support the null hypothesis when glutathione is included in the analysis. This is an indication that the intercept of glutathione is different from that of uric and ascorbic acid, which is reasonable since the thiol group can scavenge free radical more effectively [19]. The predicted IC50 values (from the least squares liner t model) are also supportive to this conclusion. (Table 2). 4. Conclusions The merging of luminol-H2 O2 -Co(II)/EDTA system to a FIA manifold enabled the provision of a sensitive method for assaying the scavenging activity of antioxidants on free radicals. The experimental set-up is designed to maximize the reaction of antioxidants with free radicals while diminish their reaction with oxidants thus improving the sensitivity and the reproducibility. Compared to routine luminometer method, the proposed method has the ability to measure the generation of oxygen-derived free radicals and antioxidants continuously in a ow system, which is closer to in vivo conditions. The utility of

the method for the quantitative determination of antioxidants in cell-free systems has also been demonstrated offering detection limits at the nmole L1 levels. F-test analysis in the parameters of the calibration curves revealed that there are no signicant differences in the reaction kinetics of these three antioxidants, which can provide a universal method for evaluating total radical trapping potential (TRAP) in biological samples or enable the determination of antioxidants scavenging activity on principal reactive oxygen species of physiological relevance. Appendix A Mechanism of luminol oxidation by H2 O2/ Co(II)/EDTA Cobalt(II) ion, as well as other transition metal ions, can increase the CL-signal of luminol-H2 O2 system, even at very low concentrations. According to Fig. 5, the abundance of Cobalt(II) ions react with H2 O2 , forming a highly reactive hydroxyl freeradical ux. At basic pH, monodissociate luminol reacts with hydroxyl radicals to water and diazamiquinone radical [26]. The latter reduces O2 to superoxide anion (O2 ) and is oxided to 5-aminohyphenthalazine-1,4dione. Diazamiquinone radical and superoxide anion (O2 ) yield the carbon-centred hydroperoxide anion that rearranges to a transient endoperoxide, which decomposes to give light emission and products, an aminophthalate and N2 . Oxygencentred radicals, such as the hydroxyl and alkoxyl radicals formed by homolytic scission of hydroperoxide, also cause photoemissive luminol oxidation. In this system, antioxidants react with hydroxyl radicals, superoxide anion and endo-peroxide attenuating the light emitted during luminol oxidation. Thus, the reaction of oxidantantioxidant becomes trivial which improves reproducibility since signal inhibition is attributed mainly to radical scavenging activity. Beyond the high reactivity of antioxidants with free

D.L. Giokas et al. / Analytica Chimica Acta 589 (2007) 5965

65

radicals, the high sensitivity (CL-inhibition) of the proposed reaction pattern can partially be attributed to interference of the endo-peroxide anion formation (due to scavenging of superoxide anions) that would otherwise decompose to give light. References
[1] K. Hensley, K.A. Robinson, S.P. Gabbita, S. Salsman, R.A. Floyd, Free. Radical. Bio. Med. 28 (2000) 1456. [2] R. Mittler, Trend. Plant. Sci. 7 (2002) 405. [3] C.L. Fattman, L.M. Schaefer, T.D. Oury, Free. Radical. Bio. Med. 35 (2003) 236. [4] M.L. Urso, P.M. Clarkson, Toxicology 189 (2003) 41. [5] J. Takata, K. Matsunaga, Y. Karube, Toxicology 180 (2002) 183. [6] A. Ghiselli, M. Serani, F. Natella, C. Scaccini, Free Radic. Bio. Med. 29 (2000) 1106. [7] J.J. Hageman, A. Bast, N.P.E. Vermeulen, Chem.-Biol. Interact. 82 (2001) 243. [8] S. Ashida, S. Okazaki, W. Tsuzuki, T. Suzuki, Anal. Sci. 7 (1991) 93. [9] O. Cynshi, Y. Takashima, Y. Katoh, K. Tamura, M. Sato, Y. Fujita, J. Biolumin. Chemilumin. 10 (1995) 261. [10] A. Dapkevicius, T.A. van Beek, H.A.G. Niederlander, A. de Groot, Anal. Chem. 71 (1999) 736. [11] E. Lissi, M. Salim-Hanna, C. Pascual, M.D. del Casillo, Free Radic. Biol. Med. 18 (1995) 153. [12] D. Mantle, J.G. Anderton, G. Falkous, M. Barnes, p. Jones, E.K. Perry, Compar. Biochem. Physiol. B-Biochem. Mol. Biol. 121 (1998) 385.

[13] A. Ogawa, H. Arai, H. Tanizawa, T. Miyahara, T. Toyooka, Anal. Chim. Acta 383 (1999) 221. [14] J.R. Trevithick, T. Dzialoszynski, Biochem. Mol. Biol. Int. 41 (1997) 695. [15] K. Van Dyke, M. Sacks, N. Oazi, J. Biolumin. Chemilumin. 13 (1998) 339. [16] T.P. Whitehead, G.H.G. Thorpe, S.R.J. Maxwell, Anal. Chim. Acta 266 (1992) 265. [17] P. Evelson, M. Travacio, M. Repetto, J. Escobar, S. Llesuy, E.A. Lissi, Arch. Biochem. Biophys. 388 (2001) 261. [18] I. Parejo, C. Codina, C. Petrakis, P. Kefalas, J. Pharmaco. Toxicol. Method. 44 (2000) 507. [19] M. Sariahmetoglu, R.A. Wheatley, I. Cakici, I. Kanzik, A. Townshend, Anal. Lett. 36 (2003) 749. [20] G. Yildiz, A.T. Demiry urek, J. Pharmaco. Toxicol. Method 39 (1998) 179. [21] I. Parejo, C. Petrakis, P. Kefalas, J. Pharmaco. Toxicol. Method. 43 (2000) 183. [22] N.P. Evmiridis, N.K. Thanasoulias, A.G. Vlessidis, Talanta 46 (1998) 179. [23] A.I. Vogel, in: G.H. Jeffery, J. Bassett, J. Mendham, R.C. Denney (Eds.), Vogels Textbook of Quantitative Chemical Analysis, fth ed., Longman Scientic & Technical, Harlow, UK, 1989, pp. 56 and 310. [24] A. Dapkevicius, T.A. van Beek, H.A.G. Niederla1nder, A. de Groot, Anal. Chem. 71 (1999) 736. [25] G.Z. Tsogas, D.L. Giokas, A.G. Vlessidis, N.P. Evmiridis, J. Chromatogr. A. 1107 (2006) 208. [26] T. Miyazawa, T. Suzuki, K. Fujimoto, K. Yasuda, J. Lipid Res. 33 (1992) 1051.