Anal Bioanal Chem (2004) 378 : 987–994 DOI 10.1007/s00216-003-2288-0
O R I G I N A L PA P E R
Roberta Andreoli · Paola Manini · Diana Poli · Enrico Bergamaschi · Antonio Mutti · Wilfried M. A. Niessen
Development of a simplified method for the simultaneous determination of retinol, α-tocopherol, and β-carotene in serum by liquid chromatography–tandem mass spectrometry with atmospheric pressure chemical ionization Received: 25 July 2003 / Revised: 4 September 2003 / Accepted: 11 September 2003 / Published online: 4 November 2003 © Springer-Verlag 2003
Abstract A new and simple method for the determination of fat-soluble vitamins (retinol, α-tocopherol, and β-carotene) in human serum was developed and validated by using liquid chromatography–tandem mass spectrometry with atmospheric pressure chemical ionization (LC-APCI-MS-MS). Different solvent mixtures were tested to obtain deproteinization and extraction of the analytes from the matrix. As a result, a volume of 240 µL of a 1:1 (v/v) ethanol/ ethyl acetate mixture added to 60 µL of serum was found to be suitable for both protein precipitation and antioxidants solubilization, giving the best recovery for all three analytes. Deproteinized samples (20 µL) were injected after dilution, without the need for concentration or evaporation to dryness and reconstruction of the sample. Vitamins were separated on a C-8 column using a 95:5 (v/v) methanol/ dichloromethane mixture and ionized in the positive-ion mode; detection was performed in the selected-reaction monitoring mode. Linearity of the LC-APCI-MS-MS method was established over 5 orders of magnitude for retinol and α-tocopherol, whereas in the case of β-carotene it was limited to 4 orders. Lower limits of quantitation were 1.7, 2.3, and 4.1 nM for retinol, α-tocopherol, and β-carotene, respectively. Serum concentrations of retinol, α-tocopherol, and α+β-carotene determined in a group of healthy volunteers were 2.48, 38.07, and 0.50 µM, respectively, in samples collected in winter (n=122) and 2.69, 45.88, and 0.90 µM during summer (n=66).
R. Andreoli · P. Manini · D. Poli ISPESL Research Center, University of Parma, Parma, Italy R. Andreoli · P. Manini (✉) · D. Poli · E. Bergamaschi · A. Mutti Laboratory of Industrial Toxicology, Department of Clinical Medicine, Nephrology and Health Sciences, University of Parma, via Gramsci 14, 43100 Parma, Italy e-mail:
[email protected] W. M. A. Niessen hyphen MassSpec Consultancy, De Wetstraat 8, 2332 XT Leiden, The Netherlands
Keywords Retinol · α-Tocopherol · β-Carotene · Serum · LC-APCI-MS-MS · Ozone
Introduction Over the last decade, there has been increased interest in the evaluation of antioxidant status in blood, since this mirrors the antioxidant defenses in target organs and tissues. For instance, the lung is subjected to oxidative damage from a variety of oxygen radicals, by-products, and oxidant gases. Antioxidant defenses in the lung are provided by endogenous enzyme systems and by dietary antioxidants, particularly vitamin C and E. The fat-soluble vitamin E and the water-soluble vitamin C are thought to act cooperatively in a system whereby cell-membranebound vitamin E is maintained in a reduced state by interaction with vitamin C [1]. An imbalance between oxidants and antioxidants in favor of the oxidant is thought to play a potential role in the pathogenesis of airway obstruction [2, 3]. Retinol (vitamin A) and its precursor (β-carotene) are known to have anti-inflammatory and antioxidant activity [4], playing a protective role against development of lung cancer and other respiratory diseases [5]. Recent studies showed the protective effect of antioxidant supplementation in workers exposed to high levels of environmental ozone [6] and the existence of a positive correlation between serum levels of antioxidants (vitamins C, E, and A, and carotenoids) and the pulmonary function in the general population [7]. In such epidemiological studies, relying on the investigation of several parameters in hundreds of subjects, analytical variability should be minimized to appreciate biological variability. From the analytical point of view, it means that accurate, sensitive, and fast methods for the simultaneous determination of many analytes in biological matrices are needed. Several methods based on the use of high-performance liquid chromatography (HPLC) have been developed to improve both the reliability and the sample throughput of the determination of blood levels of fat-soluble vitamins [8, 9, 10, 11, 12, 13]. All the published procedures for the
988
extraction of fat-soluble vitamins from serum or plasma share the same general approach. The first step, that is, protein denaturation, performed with or without saponification of the sample and with or without the addition of a stabilizing agent, is obtained by adding ethanol [8, 11, 12] or a mixture of ethyl acetate and butanol [13]. After centrifugation, the sample is extracted twice (or more) with another organic solvent (e.g., n-hexane), the organic phase is evaporated to dryness with an inert gas, and the residue is resuspended in the mobile phase used for reversedphase HPLC. The main drawback of such a procedure is that it is unsuitable for automation, even if the evaporation step could be avoided [13]. Most of the published HPLC methods have been developed for spectrophotometric or electrochemical detection. Recent papers proposing the use of mass spectrometry (MS) for the detection of antioxidant vitamins in biological fluids [14, 15, 16, 17] are mainly concerned with the detection rather than with the sample preparation. Considering the improvement in sensitivity and selectivity introduced by hyphenated techniques like MS(-MS), we developed a new and simplified analytical method for the LC-MS-MS simultaneous determination of fat-soluble vitamins in serum. This method, which requires a reduced number of analytical steps in sample handling by eliminating the unnecessary ones and merging more steps into one, could be useful not only to shorten the analysis time but also to reduce possible losses of the light-sensitive antioxidants.
Material and methods Chemicals and standards all-trans-Retinol (vitamin A, >99% pure), retinyl acetate (vitamin A acetate, internal standard, I.S.), retinyl palmitate (vitamin A palmitate, I.S., >85%), α-tocopherol (vitamin E, >98%), and α-tocopherol acetate (vitamin E acetate, I.S., >98%) were obtained from Fluka (Milan, Italy). trans-β-Carotene (95%) and butylated hydroxytoluene (BHT, 99%) were purchased from Sigma–Aldrich (Milan, Italy). All chemicals were used without further purification. HPLC-grade water, methanol, acetonitrile, ethyl acetate, and dichloromethane were supplied by LabScan (Dublin, Ireland). Analytical-grade ethanol absolute, n-hexane, and diethyl ether were from Carlo Erba Reagenti (Milan, Italy). Stock solutions of each antioxidant (approximately 10 mM) were prepared weekly in dark vials by using different organic solvents with added 0.01% (w/v) BHT (i.e., methanol for retinol, diethyl ether for α-tocopherol, and 1:1 (v/v) dichloromethane/diethyl ether for β-carotene) and were stored under nitrogen at –20°C. Stock solutions of I.S. were prepared in diethyl ether. A standard solution containing all the analytes at the concentration of 1 mM was prepared in a 1:1 (v/v) ethanol/ethyl acetate mixture, from which calibrating standards (n=18) were prepared by subsequent dilution in the same solvent mixture. A volume of 20 µL of each standard was injected in triplicate onto the LC-APCI-MS-MS system to study the linear dynamic range. Subjects and study design About 120 healthy, non-smoking volunteers (89 males and 33 females, aged 37.50±6.75 years) were recruited to participate in the study, which was approved by the Ethical Committee of the University. Venous blood samples were drawn in the morning, between
8:00 and 10:00 a.m. during fasting. All the volunteers were evaluated during winter, that is, outside from episodes of photochemical pollution (ozone peak concentrations, 11–66 µg m–3); about 50% of them (48 males and 18 females, aged 37.89±6.51) were also evaluated during summer in the morning and in the evening, after exposure to environmental ozone (81–188 µg m–3). All subjects were asked to complete a detailed questionnaire on dietary habits. Blood samples were centrifuged (2,500 g for 15 min) within 2 h after collection, and resulting serum samples were stored in dark vials at –80°C until analysis. Optimization of the sample preparation conditions The standard addition method was applied to optimize sample preparation. About 2.5 mL of a pooled normal human serum sample were used for this experiment. For the standard addition to serum samples, three mixtures of vitamin standards were prepared in methanol containing 7, 100, and 2 µM (level 1); 38, 600, and 11 µM (level 2); and 70, 1,000, and 20 µM (level 3) of retinol, α-tocopherol, and β-carotene, respectively, and a constant concentration of each internal standard (30, 400, and 10 µM of retinyl acetate, α-tocopherol acetate, and retinyl palmitate, respectively). Methanol was used to minimize serum denaturation. Addition of standards was done before denaturation of proteins (“pre” mode). Four aliquots (360 µL) were obtained from the pooled serum sample; three of these were spiked with 40 µL of the standard mixtures and the fourth with the same volume of methanol. Each aliquot was divided again into five aliquots (60 µL) and treated according to the different denaturation–extraction procedures indicated in Table 1. After addition of the extraction solvent, the samples were vortexed and centrifuged at 11,000 rpm for 4 min. In order to establish whether protein precipitation could influence vitamin levels, addition of standards was also done after denaturation of proteins (“post” mode). In this case, five aliquots of each serum sample were added with the deproteinizing solvents reported in Table 1, vortexed, and centrifuged. Then, denatured samples were divided into four aliquots, which were extracted with the different solvents indicated in Table 1 containing standard mixtures of vitamins in order to obtain the same final concentrations used in the “pre” mode. Sample preparation for the determination of fat-soluble vitamins in serum samples Method C (Table 1, “pre”) was used for sample analysis. Serum samples (60 µL) were denatured with 240 µL of a 1:1 (v/v) ethanol/ ethyl acetate mixture, vortexed, and then centrifuged at 11,000 rpm for 4 min. Then, 20 µL of the denatured sample was injected twice onto the chromatographic system. Due to the large number of samples and to the impossibility of applying the standard addition method to every sample, quantitative analysis was performed using an external calibration. In order to minimize the matrix effect, calibration curves were obtained by spiking a pooled serum sample, whose analyte concentration was previously determined by the standard addition method, with fresh standard mixtures at four different concentrations in the ranges 0–10, 0–100, and 0–2 µM for
Table 1 Scheme of the deproteinization–extraction procedures applied to serum samples for method optimization Method
Serum (v)
Denaturating solvent (v)
Extracting solvent (v)
A B C D E
1 1 1 1 1
2 AcCN 2 EtOH 2 EtOH 2 EtOH 2 EtOH
2 AcCN 2 EtOH 2 AcEt 2 CH2Cl2 2 n-Hexane
989 retinol, α-tocopherol, and β-carotene, respectively, and denatured as samples. Calibration curves were constructed by linear regression analysis of the areas of each analyte versus the concentration (µM) injected, which was calculated as the sum of the natural content and the concentration spiked. A whole calibration set was run each 20 samples, whereas every tenth analysis one of the calibration standard was injected to verify the stability of the analytes and the reproducibility of the LC-APCI-MS-MS system. Each sample was injected twice. Liquid chromatography–tandem mass spectrometry LC-MS-MS analyses were carried out by using an Applied Biosystems-Sciex API 365 triple-quadrupole mass spectrometer (Sciex, Concord, Canada) equipped with an atmospheric pressure ionization (API) source and a heated nebulizer APCI interface. A Power Macintosh G3 computer was used for instrument control, data acquisition, and processing. The liquid chromatograph was a Perkin– Elmer series 200 binary system (Norwalk, CT, USA) equipped with an ASPEC XL autosampler (Gilson, Villiers-le-Bel, France). Antioxidants were analyzed in the same chromatographic run by using a Supelcosil LC-8-DB column (150 mm×4.6-mm I.D., 3 µm; Supelco, Bellefonte, PA, USA). Isocratic elution of the analytes was performed with a methanol/dichloromethane mixture (95:5, v/v) at a flow rate of 0.80 mL min–1. The analytes were ionized by APCI in positive-ion (PI) mode. Final ionization conditions were needle current 3 µA, heated nebulizer temperature 400°C, curtain gas nitrogen (0.95 L min–1). Particulate-free and CO2-free air was used as the nebulizing gas at a flow rate of 1.23 L min–1. The standard solutions (25 µM) used for the parameters optimization were in ethanol/ethyl acetate (1:1 v/v). The product-ion spectra were obtained by selecting the [M+H–18]+ ion for retinol, the [M]· for α-tocopherol, and the [M+H]+ for β-carotene, respectively; the corresponding scan ranges were m/z 100–300, 100–450, and 100–580. Quantitative determinations were obtained in selected-reaction monitoring (SRM) mode, following the reactions m/z 269→213 (collision energy 17 eV) for retinol, retinyl acetate, and retinyl palmitate; m/z 430→165 (29 eV) for α-tocopherol; m/z 473→207 (29 eV) for α-tocopherol acetate; m/z 537→177 (21 eV) for β-carotene. Instrumental control was achieved by the Sample Control software version 1.4, whereas for peak integration and quantitation the MacQuan 1.6 software was used. Statistics Statistical analysis was carried out by using the SPSS/PC+ software (Version 10.0 for Windows, Chicago, IL, USA). Linear regression analysis (Pearson’s correlation) was used to construct calibration curves for each analyte. Parametric statistical tests were applied to log-transformed values, when necessary to obtain a normal distribution, which was assessed by using the one-sample Kolmogorov–Smirnov test. Values were expressed as geometric mean (GM) and geometric standard deviations (GSDs). Differences between groups were primarily assessed by using the t-test for paired samples. A p value of less than 0.05 indicated statistical significance.
cule [MH–H2O]+ (m/z 269); no intact protonated molecule could be observed. Also in the case of retinyl esters, the base peak was detected at m/z 269 and was formed by the elimination of acetic acid and palmitic acid in the APCI ion source [14]. Both the molecular ion [M]· (m/z 430) and the protonated molecule [M+H]+ (m/z 431) in a constant ratio were observed for α-tocopherol, whereas for β-carotene only the [M+H]+ (m/z 537) was present. The formation of the α-tocopherol molecular ion next to the protonated molecule was previously observed by Lauridsen et al. [15]. The same authors also noticed the presence of an additional ion at m/z 429 (not observed in our specta), which after high-resolution measurements on an FT-MS instrument was attributed to [M–H]+ formed by initial protonation of α-tocopherol followed by dehydrogenation. APCI plasma is known to generate protonated molecules for carotenes, including β-carotene [18], whereas in CI, DCI, and ESI the molecular ion [M]· is generated [19]. Production mass spectra of the three antioxidants are shown in Fig. 1. Except for α-tocopherol, APCI-MS-MS characterization of these compounds has never been reported before. Fragmentation of retinol leads to the formation of a fragment ion at m/z 213, attributable to the loss of 56 Da (presumed rearrangement and neutral loss of butene) from dehydrated retinol. As reported [15], the spectrum of α-tocopherol shows a unique and intense product ion at m/z 165 similarly to in EI ionization [20]. Fragmentation of β-carotene gives rise to a complex product ion spectrum, typical of conjugated polyolefins. A pattern of structurally significant fragment ions is evident at m/z 137, 177, 203, 243, and 269, presumably due to protonation followed by breaking in correspondence with the C7–C8, C9–C10, C11–C12, C13–C14, and C15–C15′ double bonds, respectively, with charge retention on the lower C number. The most intense ion of this series (m/z 177) is probably stabilized by the formation of a ring structure. The ions at m/z 123, 189, 255, 281, 321, 347, and 413 belong to another series generated by the cleavage of the intermediate single bonds with retention of the charge possible on either sides, that is, C6–C7 (m/z 123 or 413), C10–C11 (m/z 189 or 347), C12–C13 (m/z 321), C14–C15 (m/z 255 or 281). The fragment ions observed in APCI-MS-MS are different from those obtained in the case of positive ion CID APCI of α-carotene [18]. Development of a new sample preparation procedure for fat-soluble vitamins in serum
Results and discussion Mass spectrometry As already observed by other authors, retinol [14], α-tocopherol [15], and β-carotene [16, 17, 18] are efficiently ionized by PI APCI. Moreover, it has been reported that APCI-MS showed a wider dynamic range and linearity of detector response in the case of retinol [14] and β-carotene [17] than with electrospray MS. Ionization of retinol occurs via protonation and immediate loss of a water mole-
The aim of this preliminary experiment was the optimization of a simplified, ideally one-step, sample preparation procedure, which allowed high recovery and reproducibility for all three analytes. Due to its complexity, serum cannot be injected without prior treatment with a solvent or a solvent mixture. The ideal solvent (mixture) should have the following characteristics: (i) allow protein denaturation; (ii) have a sufficient apolarity and a good affinity for all three analytes to avoid co-precipitation of vitamins during protein denaturation; (iii) be completely miscible
990 Fig. 1a–c Flow-injection PI-APCI-MS-MS product-ion mass spectra of a retinol, precursor ion [MH–H2O]+, m/z 269, b α-tocopherol, precursor ion [M]·, m/z 430, and c β-carotene, precursor ion [M+H]+, m/z 537
with the sample to avoid separation of phases and the need for an authentic extraction; (iv) be compatible with the detection system (i.e., not to suppress APCI ionization). All these parameters have been evaluated in a preliminary set of experiments. Two of the most widely used deproteinizing solvents, acetonitrile and ethanol, used alone or mixed with other solvents of decreasing polarity (ethyl acetate, dichloromethane, and n-hexane), were compared in this study. The addition of another less polar organic solvent could be important to increase the solubility of the analytes in the final mixture and to avoid their loss during denaturation. Table 1 summarizes the scheme of sample pre-treatment followed in the optimization of the sample preparation procedure. Since a strong ionization suppression was observed for α-tocopherol (10-fold) in n-hexane and for β-carotene (5-fold) using dichloromethane, stan-
dards for calibration were prepared in the matrix (a pooled serum sample) and treated exactly like samples. Moreover, the addition of standards to serum samples was done both before (“pre” mode) and after (“post” mode) denaturation of proteins, with the aim of verifying possible matrix effects, such as co-precipitation of vitamins during the deproteinization step. Figure 2 compares the vitamin natural content of the pooled serum sample, obtained by applying the different sample preparation methods (A–E) described in Table 1, and recalculated by means of the standard addition method in both “pre” and “post” mode. This procedure should correct for both the solvent and the matrix effect, leading to similar pre/post values for all methods. Nevertheless, in most cases this correction was not adequate, as shown by Fig. 2. Only the use of an ethanol/ethyl acetate mixture
991 Table 2 Reproducibility of methods A–E, evaluated on 12 independent extractions of the same serum sample. Values are expressed as %RSDs Analyte
I.S.
Methods A
B
C
D
E
Retinol
yes no
56 24
21 6.4
11 9.4
41 48
26 19
α-Tocopherol
yes no
33 47
3.0 5.8
7.1 3.6
20 54
16 20
β-Carotene
yes no
25 55
3.9 5.7
43 75
14 26
25 22
ods D and E). For the most polar solvents (acetonitrile and ethanol used alone), the difference in the pre/post concentrations of α-tocopherol and β-carotene (cpre–cpost) could be explained by the scarce solubility and the variable extent of co-precipitation of the analytes during protein denaturation. By contrast, the addition of apolar solvents to denatured serum was found to cause a lack of miscibility and extremely poor reproducibility (dichloromethane), or separation of phases and partition of analytes (n-hexane). Since a single extraction with n-hexane was not exhaustive, a variable percentage of the analytes remained in the ethanolic phase, making method E unsuitable for the scope of this work. The reproducibility of different methods, evaluated by means of 12 independent precipitations/extractions of the same serum sample, is described in Table 2. This data further support the conclusion that only the ethanol/ethyl acetate mixture (method C) has the right characteristics to be used in a rapid, one-step sample preparation procedure for vitamin LC-MS analysis, leading also to sufficiently reproducible results. The recovery of method C was calculated by spiking a serum sample, whose vitamin concentrations were previously measured, with standard mixtures of known concentrations (1-, 2.5-, and 5-fold the natural content). Recoveries (mean±SD, n=12) were 98.0±1.9% for retinol, 100.8±1.7% for α-tocopherol, and 95.0±2.7% for β-carotene, respectively.
Fig. 2 Comparison among the vitamin natural contents of a pooled serum sample obtained by applying different sample pretreatment methods (A–E) described in Table 1 and the standard addition method. Methods: A AcCN; B EtOH; C EtOH/AcEt; D EtOH/ CH2Cl2; E EtOH/n-hexane. Addition of standards was done before serum denaturation (“pre” mode) and on denatured serum (“post” mode). *Pre/post concentrations not statistically different (p>0.05)
(method C) gave non-statistically different pre/post concentrations (p>0.05) for all antioxidants, thus indicating a good affinity of the mixture for all analytes and no matrix effects. For other methods, despite the use of I.S., the simultaneous presence of both the matrix and the solvent (mixture) probably resulted in a non-homogeneous system, in which the spiked standard and the natural analyte content behaved differently with respect to co-precipitation (methods A and B) and ionization suppression (meth-
Validation of the LC-APCI-MS-MS method Figure 3 shows the LC-APCI-SRM analysis of vitamins in an authentic serum sample treated according to method C. As all the analytes are isocratically eluted within 6.5 min, the method is suitable for high-throughput routine analysis. By using a C-8 column, β-carotene (retention time 6.1 min) could be separated from other isomeric carotenoids present in serum, for example, lycopene (acyclic, 5.0 min) and γ-carotene (monocyclic, 5.5 min), but not from α-carotene (bicyclic-like β-carotene). According to other authors [11, 12, 13, 16, 21], α-carotene and β-carotene could be separated using a C-18 column. By using the same eluent, a 95:5 (v/v) methanol/dichloromethane mixture, separation was achieved in 15 min (data not shown). It should be noted that interference from lycopene in the determination of β-carotene could be excluded by using MS/MS
992 Fig. 3a–d LC-PI-APCI-MS-MS chromatogram obtained in SRM mode of an authentic human serum sample from a healthy volunteer spiked with I.S. Chromatographic conditions: column Supelcosil LC-8-DB; mobile phase methanol/ dichloromethane (95:5 v/v); flow rate 0.80 mL min–1. SRM transitions and peak identification: a m/z 269→213 (17 eV) retinol (1), retinyl acetate (I.S., 2), retinyl palmitate (I.S., 3); b m/z 430→165 (29 eV) α-tocopherol (4); c m/z 473→207 (29 eV) α-tocopherol acetate (I.S., 5); d m/z 537→177 (21 eV) α+β-carotene (6)
also in case of coelution. In fact, the fragment ion chosen for quantitation (m/z 177) contains the cyclic terminal isoprene and is not formed by the acyclic lycopene, which has been reported not to fragment in PI-APCI MS-MS [21]. Since the method does not require any authentic extraction (not resulting into a phase separation) or reduction of volume, but it implies simply a 1:5 dilution followed by direct injection, the use of I.S. was found not to improve the reproducibility of the results but conversely to increase their variability (see Table 2). This was probably due to the fact that analogous molecules (esters of vitamins) rather than isotopically labeled compounds were used as I.S. in this study. Moreover, retinyl and α-tocopherol esters are already present in some blood samples probably due to the consumption of vitamin-fortified food. For these reasons, we decided to perform quantitative analy-
sis of samples without the addition of I.S. The LC-APCIMS-MS method was finally validated by studying the linear dynamic range, the limits of detection (LODs), the lower limits of quantitation (LLOQs), and the intra-day and inter-day precisions for each antioxidant. Results are summarized in Table 3. The LODs, calculated as the ratio S/N>3 and obtained in SRM mode, were at least 10- to 20-fold better than those reported in the literature for retinol [14], α-tocopherol [17], and β-carotene [16] using MS detection. The high sensitivity of the method allowed the injection of diluted samples (1:5) without the need for evaporation to a small volume or even to dryness for all the examined analytes including retinol, for which the lowest basal concentrations are reported.
993 Table 3 Linear dynamic range, correlation coefficients (r2), limits of detection (LODs), lower limits of quantitation (LLOQs), and precision of the LC-APCI-MS-MS method for the determination of antioxidants in seruma Compound
Range
a×10–4b
b×10–5b
r2
n
LOD (nM)c
LLOQ (nM)d
%RSD intra-day
%RSD inter-day
Retinol α-Tocopherol β-Carotene
0.003–350 0.002–950 0.009–750
4.23±0.03 7.65±0.13 6.32±0.32
1.56±4.99 9.05±3.25 –0.61±5.67
0.998 0.997 0.998
48 57 39
0.7 1.2 1.9
3.9 2.3 4.1
3.9 2.1 4.7
5.7 4.5 7.3
aCalibration
dLower
b±Values
ditions
curve: y=ax+b are confidence intervals at 95% probability level cLimit of detection (S/N=3) calculated under SRM conditions
Application
Conclusions
One of the aims of this study was to investigate the possible correlation between blood vitamin content and environmental ozone exposure. The LC-APCI-MS-MS method was applied to the determination of serum concentrations of vitamins in a group of healthy subjects. To assess possible changes of blood concentrations of antioxidants, the subjects were evaluated both in a period without photochemical smog (i.e., during winter) to determine the basal antioxidant status and during a period characterized by relatively high concentrations of ozone (i.e., during summer). The results obtained, expressed as geometric means (GM) and geometric standard deviations (GSD), for the subjects are summarized in Table 4. Antioxidant concentrations are consistent with those reported in other studies [7, 8, 11]. Although exposure to ozone apparently did not affect the serum levels of antioxidants, differences between winter and summer levels of vitamins are observable for α+β-carotene and α-tocopherol (p
