Second-order calibration of excitation–emission matrix fluorescence spectra for determination of glutathione in human plasma

Talanta 79 (2009) 648–656

Contents lists available at ScienceDirect

Talanta journal homepage: www.elsevier.com/locate/talanta

Second-order calibration of excitation–emission matrix fluorescence spectra for determination of glutathione in human plasma Bahram Hemmateenejad a,∗ , Zahra Rezaei b , Sasan Zaeri b a b

Department of Chemistry, Shiraz University, Shiraz 71454, Iran Department of Medicinal Chemistry, Faculty of Pharmacy, Shiraz University of Medical Sciences, Shiraz, Iran

a r t i c l e

i n f o

Article history: Received 24 January 2009 Received in revised form 18 April 2009 Accepted 21 April 2009 Available online 3 May 2009 Keywords: Glutathione Fluorescence Excitation–emission Second-order calibration Standard addition Plasma

a b s t r a c t A rapid non-separative spectroflourimetric method based on the second-order calibration of the excitation–emission data matrix was proposed for the determination of glutathione (GSH) in human plasma. In the phosphate buffer solution of pH 8.0 GSH reacts with ortho-phthaldehyde (OPA) to yield a fluorescent adduct with maximum fluorescence intensity at about 420 nm. To handle the interfering effects of the OPA adducts with aminothiols other than GSH in plasma as well as intrinsic fluorescence of human plasma, parallel factor (PARAFAC) analysis as an efficient three-way calibration method was employed. In addition, to model the indirect interfering effect of the plasma matrix, PARAFAC was coupled with standard addition method. The two-component PARAFAC modeling of the excitation–emission matrix fluorescence spectra accurately resolved the excitation and emission spectra of GSH, plasma (or plasma constituents). The concentration-related PARAFAC score of GSH represented a linear correlation with the concentration of added GSH, similar to that is obtained in simple standard addition method. Using this standard addition curve, the GSH level in plasma was found to be 6.10 ± 1.37 ␮mol L−1 . The accuracy of the method was investigated by analysis of the plasma samples spiked with 1.0 ␮mol L−1 of GSH and a recovery of 97.5% was obtained. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Glutathione (GSH) is a tri-peptide consisting of three amino acids: glutamate, cysteine and glycine. It has a key role in the defense system of many mammalian tissues. A look at the literature will introduce us to an extensive variety of functions attributed to GSH among which antioxidant activity against both endogenous and exogenous reactive species outstands [1,2]. GSH can be considered as a redox buffer [3,4]; it scavenges free radicals and detoxifies xenobiotics and is a strong defense against oxidative and nitrosative stresses [5–7]. GSH has also role in regulating cell proliferation [8] as well as cell apoptosis, the programmed death of cells [9–11]. Glutathione has two main redox forms; GSH is the reduced form of glutathione and is actually the predominant form of it over other oxidized forms [12,13]. Among the oxidized forms, GSSG is known more and is both enzymatically and non-enzymatically formed by GSH [14]. In most of the cells, GSSG accounts for only 1% of the depleted GSH whereas it comprises more than 25% of the oxidized GSH in the subcellular parts, e.g. in the mitochondria [15,16]. For some reasons, reported reference levels of glutathione seem to vary from one work to another in the literature. These differ-

∗ Corresponding author. Tel.: +98 711 2284822; fax: +98 711 2286008. E-mail address: [email protected] (B. Hemmateenejad). 0039-9140/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.talanta.2009.04.041

ences are mainly derived from different methodologies adopted [17]. So, a good sample preparation including, for example, avoidance of unwanted hemolysis of the red blood cells (RBCs) or the storage of whole blood or plasma samples under an appropriate temperature condition [18] is of crucial importance in the assay of glutathione. In almost all of the methods for glutathione determination, a deproteinization step is necessary [14]. Apart from that, such methods mostly adopt a derivatizing agent that can be either colorimetric or fluoregenic. Sometimes when the determination of total glutathione is desired, a disulfide reduction step is also included [19]. Therefore, all these types of sample interventions can be responsible for the variations seen in the determination results. Such fluctuations from one work to another may be well recognized when blood and plasma samples are concerned; and out of these two assay environments, plasma data show more diversity; maybe because plasma undergoes more manipulations than whole blood in the preparation step [17]. There are different methodologies that are adopted in the quantification of glutathione and its analogues. These methods are classified mainly as separative and non-separative techniques. Chromatographic approaches including high performance liquid chromatography (HPLC), thin layer chromatography (TLC) and gas chromatography (GC), together with capillary electrophoresis (CE) serve as separative ones while spectrophotometric, spectrofluorimetric and electrochemical methods fall in the second category

B. Hemmateenejad et al. / Talanta 79 (2009) 648–656

[14]. Chromatographic techniques take the advantage of high specificity and selectivity for they physically neutralize the interfering effects of artifacts and when used in combination with sensitive detectors like fluorimetric ones, they become so sensitive and that is why there have been so increasing attention towards them. On the other hand, non-separative approaches are not very well reflected in the works of recent investigators mainly because of the problem of unsatisfactory specificity caused by other aminothiols present in the biological matrices. Although non-separative techniques lack enough specificity, they are very promising with respect to the matter of their simplicity and low cost if we manipulate the assay system in a way so as to somehow separate and probably recognize the interferences contributing to the final instrumental signals. GSH and its analogues do not present any strong chromophores or fluorophores in their structures and hence, are frequently derivatized by either colorometric or fluoregenic tags to exhibit convincing sensitivity [20]. Except for the electrochemical-based detection methods which do not rely on the derivatization of glutathione, other methods lack enough sensitivity when no derivatization has been considered and therefore have either a chromophore or fluorophore introduction step in their protocols [14]. Fluoregenic labeling agents are commonly used in the assay of glutathione. They offer high sensitivity compared to UV–vis labeling agents. These fluoregenic agents are o-phthaldehyde (OPA) [21–25], monobromobimane (BrB) [26], fluorobenzofurazan derivatives [27,28], Rhodamine-based probes [29,30], etc., among which OPA (not fluorescent by itself) is widely used to form a highly fluorescent, stable derivative of GSH and other aminothiols. The mild laboratory conditions required for the reaction of OPA with GSH have made this reagent much attractive to be used by many investigators involved in glutathione determination [21–25,31]. The introduction of OPA as a derivatizing agent in the assay of glutathione was done by Cohn and Lyle [23] and was then modified by other authors [24]. Although OPA serves as a good labeling tag as mentioned before, it is very active towards amino acids as well as aminothiols other than GSH, putting a challenge in front of its applications in the GSH assay [17,31–35]. Since the introduction of chemometrics methods in analytical chemistry the problem of spectral overlapping has been diminished thanks to the resolving power of various multivariate calibration methods. In contrast to univariate calibration, measuring of multivariate signals per sample enables one to compensate for contributions of interferences in an unknown sample. While the first-order multivariate calibration methods are able to handle the spectral interferences of the compounds whose variations are taking into account in the calibration process [36–41], the second- or higher order data analyses methods can compensate for potential interferences not included in the calibration set [42–45]. This is universally recognized as the second-order advantage. The secondorder methods need a matrix of response data per sample and thus a three-way array of data is obtained by staking the data matrices of different samples under each other. These types of data can be provided by hyphenated instruments such as HPLC-DAD, GC–MS and LC–MS, excitation–emission fluorescence spectra and spectroscopic monitoring of the reaction kinetic. Parallel factor (PARAFAC) analysis is one of the second-order calibration methods, in which the trilinearity of the measured analytical data is a necessary condition. For a detailed discussion on PARAFAC and its basis and applications, the reader is referred to the literature [42–47]. The excitation–emission matrix of fluorescence spectra is a kind of trilinear data and such data have been extensively used in the recent years to achieve second-order advantage. In this work, we developed a non-separative spectrofluorimetric-based method, with OPA as the labeling agent for the determination of plasma GSH using PARAFAC as a second-order data analysis method. We adopted a second-order

649

standard addition method to compensate both the contributions of other aminothiols, potentially presented in plasma, and the effect of plasma matrix on the spectrofluorimetric determination of GSH. The experimental conditions were optimized to achieve the best sensitivity for the direct determination of GSH in plasma. 2. Experimental 2.1. Reagents o-Phthaldehyde (OPA), trichloroacetic acid (TCA), Na2 HPO4 and EDTA were obtained from Merck Co. (Germany). GSH was purchased from Sigma Chemical Company (St. Louis, MO). Doubly distilled water was used to make disodium hydrogen phosphate buffer solution. The buffer pH was set at 8.0 by dissolving 1.780 g of disodium hydrogen phosphate in doubly distilled water and the final volume of it was made to 100.0 mL. Apart from disodium hydrogen phosphate, 0.1 g of EDTA was also dissolved in the buffer solution to inhibit the autoxidation phenomenon of GSH which is actually common and problematic in its assay. The mother OPA solution used for derivatization was prepared by dissolving 0.025 g OPA in 25.0 mL methanol (reagent grade, Merck) to yield a concentration of 0.1% (w/v) of OPA. This solution seemed to maintain its activity for several weeks when kept in the refrigerator. GSH stock solution was prepared by dissolving 0.015 g in 100.0 mL of phosphate buffer solution to yield a concentration of 500.0 ␮mol L−1 and stored at 4 ◦ C until used. The 10% (w/v) TCA solution was prepared by dissolving 10.0 g of cold TCA crystals in doubly distilled water and made to 100.0 mL. 2.2. Instrumentation A Perkin-Elmer LS 50B Luminescence Spectrophotometer was used for the fluorimetric measurements. A refrigerator-equipped centrifuge model SIGMA 3K30 was used for precipitating plasma proteins at high (10,000 × g) revolution. Data manipulation was performed employing Microsoft Excel (2003) and MATLAB 7.0. 2.3. Plasma samples and sample preparation Daily based fresh frozen plasma (FFP) samples prepared from the venous blood of random healthy male and female blood donors of the Central Blood Transfusion Organization (Shiraz, Iran) were gathered. The process of RBC removal and other necessary steps in preparing plasma samples were done and checked by the staff of that organization. Frozen plasma samples with a temperature lower than −80 ◦ C were put into standard cold boxes and quickly transported to the laboratory freezers until the day of experiment. As suggested by many authors that have worked on the glutathione determination, one of the necessary steps in plasma preparation is protein removal that is done by different strategies including the use of acids (trichloroacetic acid (TCA), perchloric acid (PCA), 5-sulfosalicylic acid (5-SSA) and metaphosphoric acid (MPA)), organic solvents (methanol, acetonitrile, etc.) and ultrafiltration, among which acidic reagents are, by far, more popular than others [14]. TCA with a concentration of 10% (w/v), served as the protein precipitant in our study. Before adding appropriate volumes of acid to plasma samples, it was given enough time for the frozen plasma to gradually melt at the room temperature (25 ◦ C). After centrifugation of acidic plasma samples at 10,000 × g (twice for each sample) in a refrigerator-equipped centrifuge maintained at 4 ◦ C, the supernatant liquid was carefully pipetted into laboratory tubes and frozen at −80 ◦ C before the spectrofluorimetric measurements. Such frozen supernatants are supposed to contain GSH and

650

B. Hemmateenejad et al. / Talanta 79 (2009) 648–656

other non-macromolecular plasma constituents such as ions, amino acids, some small peptides etc., but no proteins. 2.4. Excitation–emission spectrofluorimetric setup All samples were measured in a dark condition because of the GSH–OPA adduct susceptibility to exposure to the direct light. Measurements were done in a 10 mm × 10 mm quartz cuvette at the room temperature. Excessive care was taken to fully clean the cuvette to reduce any fluorescent interfering hand-transported contaminations on its external wall. The wavelength range of 370–600 nm with 0.5 nm intervals was selected for obtaining the emission spectra of the assay samples. For each sample, an excitation–emission data matrix was generated by exciting that sample at 10-nm intervals from 310 to 360 nm. It was found that the best intensities were gained if the excitation and emission slit widths were set at 10 nm and 3.5 nm, respectively. A scan speed of 1500 nm/min was employed. It should be noted that for univariate calibration the excitation and emission wavelengths were set at 340 nm and 420 nm, respectively. The data produced by the instrument were smoothed using the Savitski-Golay smoothing approach and subsequently analyzed by PARAFAC. 2.5. Calibration of the GSH–OPA adduct in buffer solution Prior to going to GSH measuring in real samples, a calibration step in the phosphate buffer medium was thought to be necessary to firstly evaluate the fluorescence of the GSH–OPA adduct as well as the appropriate excitation–emission wavelength pair reported in the literature and secondly to draw a calibration curve for the determination of the detection limit in the phosphate buffer solution. To do so, various GSH solutions with concentrations of 0.20, 1.00, 2.00, 4.00, 5.00, 10.00, 20.00 and 40.00 ␮mol L−1 were prepared in the aqueous buffer of pH 8.0. Equal volumes (i.e., 2.0 mL) of 0.1% OPA and GSH solutions of desired concentration were directly mixed in the well-wrapped laboratory tubes and after mixing gently, appropriate portion of the reaction mixture was transferred into the spectroflourimeter cuvette. It should be noted the concentrations of GSH in the resulting solutions are one-half of the above mentioned concentrations. The fluorescence intensity of each reaction solutions was then measured at the excitation and emission wavelengths of 340 nm and 420 nm, respectively. Three independent measurements were performed for each solution at the same lab conditions. Calibration curve was derived by plotting the averaged fluorescence intensity vs. GSH concentration. 2.6. General procedure for GSH assay in plasma A standard addition approach coupled with PARAFAC as threeway calibration method was adopted here for the assay of GSH in plasma. In a series of six well-wrapped laboratory tubes were added 0.50 mL diluted plasma and 3.25 mL aqueous buffer of pH 8.0. Then minute and variable amounts of GSH stock solution (i.e., 0.0, 8.0, 24.0, 40.0, 80.0 and 160.0 ␮L) were added to the above solutions to make the added GSH concentrations of 0.0, 1.0, 3.0, 5.0, 10.0 and 20.0 ␮mol L−1 , respectively. Finally, 0.25 mL of 0.10% (w/v) OPA was added to reach to the final volume of 4.00 mL. The reaction mixture was then allowed to stand in darkness for 10 min at the room temperature for the reaction to be completed. Since the plasma also contains GSH, the actual concentrations of GSH in the resulting solutions were not equal to the added concentrations. The degree of plasma dilution was optimized to obtain the best results with respect of both accuracy and sensitivity. Under the specified instrumental setup, an excitation–emission data matrix was generated for each of six calibrator samples by successively exciting each sample at 360, 350, 340, 330, 320

and 310 nm and obtaining the corresponding emission intensities through 370–600 nm. To show that the analysis of the GSH concentration in the plasma is valid, i.e., to check the model recovery power, the plasma samples were spiked with 1.0 ␮mol L−1 of standard GSH and the complete standard addition procedure discussed in the previous paragraph was employed and the percent of recovery was calculated. 2.7. PARAFAC modeling The PARAFAC analysis was performed using the N-way toolbox of MATLAB provided by Professor Rasmus Bro. It was taken from the website of Faculty of Life Sciences, University of Copenhagen (http://www.models.kvl.dk/source/). The recorded three-dimensional excitation–emission fluorescence data of standard addition samples were denoised employing Savitsky–Golay smoothing method. The six two-dimensional matrices were rearranged into a three-dimensional array required for PARAFAC analysis. Final data were the average of the three independent runs at the same experimental conditions. Core consistency, percent of the explained variances (or residual sum of square errors) and number of PARAFAC iterations were used to obtain the optimum number of PARAFAC factors. To obtain the best results, PARAFAC was run using different applied constraints. 3. Results and discussion The unique properties of GSH–OPA adduct (i.e., mild reaction condition, no fluorescence emission of OPA and intense fluorescence emission of adduct) made it as a routine detection method for glutathione by liquid chromatographic methods [31]. Thanks to the resolving power of three-way calibration methods even in the presence of non-modeled interferences, here, PARAFAC calibration was used to develop a rapid analytical method for quantization of GSH in plasma without need to separate the analyte from interferences. First of all, the fluorescence property of GSH–OPA adduct was studied in the buffer solution. It was found that the best excitation–emission wavelength pair is 340 nm and 420 nm, respectively as also suggested by the literature [21–25,48]. The excitation and emission fluorescence spectra of GSH–OPA adduct in buffer of pH 8.0 are shown in Fig. 1. The excitation and emission spectra exhibit distinct peak maxima at 340 nm and 420 nm, respectively. It should be noted that pH was not optimized in this work and the previously reported optimized value (pH 8.0) was used. By plotting emission intensities (with excitation and emission wavelengths of 340 nm and 420 nm, respectively) vs. GSH concentrations, a linear calibration curve in the GSH concentration range of 0.05–10.00 ␮mol L−1 with calibration equation of (y = 34.63 × −1.14) and correlation coefficient (R2 ) of 0.990 was obtained. The averaged emission intensity of the blank sample containing no GSH was 3.51 with a standard deviation of 0.077 (n = 5). Three times of the blank standard deviation was added to the average of blank emission intensity to obtain the detection limit (DL) signal. Accordingly, a detection limit (DL) of 0.039 ␮mol L−1 was obtained for GSH determination in buffer solution by transferring the DL signal into the calibration equation. The next step is studying the fluorescence property of GSH–OPA adduct in the human plasma solution. There are numerous known and unknown biological compounds such as SH-bearing amino acid, cysteine, and other thiol-containing peptides as well as many fluorescent compounds present in the plasma matrix that may severely interfere with the analyte under assay. The excitation and emission spectrum of plasma is also shown in Fig. 1. Plasma, when untreated with OPA, shows an excitation spectrum having maximum intensity at 280 nm and a descending emission pattern

B. Hemmateenejad et al. / Talanta 79 (2009) 648–656

651

Fig. 1. Excitation (a) and emission (b) spectra of GSH–OPA adduct and plasma (untreated with OPA).

through the wavelengths of 370–600 nm. Surprisingly, a relatively significant shoulder at around 420 nm was also observed in the plasma emission profile. In addition to fluorescence emission, plasma also imposes its quenching effects on the fluorimetric quantification of the analytes present in it. This phenomenon becomes very problematic in deriving satisfactory models because quenching effects severely reduce enough sensitivity required for the assay. In addition, the plasma constitutes can also show interfering effects through reaction with OPA. To investigate the effects of plasma on the fluorescence prop-

erty of GSH–OPA adduct, two separate experiments were run: (a) firstly, OPA and GSH solutions were mixed and after the completion of the reaction, plasma solution was added and (b) secondly GSH standard solution was added to plasma and then they were mixed with OPA solution. Both experiments were achieved with variable volumes of the plasma solution. The resulted emission spectra are shown in Fig. 2. It should be noted that in the above experiments, the concentrations of GSH and OPA were constant and only the added volumes of plasma solution in the total of 4.0 mL reaction mixtures were varied between 0.0 and 2.5 mL. Moreover, the plasma itself

Fig. 2. Quenching of GSH–OPA adduct by plasma. The order of addition of reagents is (a) OPA and GSH solutions were mixed and after the completion of the reaction, plasma solution was added, (b) firstly GSH standard solution was added to plasma and then they were mixed with OPA solution. The volumes of added plasma are between 0.0 mL (the highest spectrum) and 2.5 mL (the lowest spectrum) with 0.5 mL intervals in a total volume of 4.0 mL.

652

B. Hemmateenejad et al. / Talanta 79 (2009) 648–656

contains some GSH but at very lower level of that added to the reaction mixture so that the contribution of plasma GSH in the GSH–OPA fluorescence can be ignored. The left part of Fig. 2 shows a gradual decreasing in the fluorescence intensity of GSH–OPA adduct (at 420 nm) upon increasing addition of plasma, accompanying with the appearance of new peak at about 520 nm. The new peak can be attributed to the reaction of OPA with other plasma constituents. This is confirmed by the spectral changes shown in the right part of Fig. 2. As it is evident, by the addition of OPA to the mixture of plasma and GSH, the fluorescence intensity of GSH–OPA is decreased dramatically accompanying with a clear formation of new florescent peak at 520 nm. The substantial decreasing of the fluorescence intensity upon addition of minute amount of plasma can be attributed to two factors: consumption of OPA by plasma constitutes and quenching effect of plasma on the GSH–OPA fluorescence. 3.1. Optimization of OPA concentration The above observations indicated significant quenching effect of plasma constituents on the GSH–OPA fluorescence. The major result of this effect is a decreased sensitivity of the proposed method for

GSH assay in plasma. While PARAFAC is able to handle the nonmodeled interfering effects of the unknown plasma constituents (i.e., those produced fluorescent products upon reaction with OPA), the mentioned matrix effect cannot be resolved by PARAFAC. To decrease matrix effect, two different criteria were employed: (a) since the tendency of OPA to GSH is more than other plasma constituents [21–26], the amounts of added OPA was decreased to the levels necessary for GSH reaction in order to decrease the formation of plasma constituents–OPA adducts and hence their interfering effects and (b) according to the emission spectra shown in Fig. 2, the volume of the added plasma to the reaction mixture was decreased in order to lower the concentration of quenchers. Up to now, the reaction mixtures contained 1.50 mL of 0.10% OPA, 0.50 mL plasma and 2.00 mL aqueous buffer to reach to the total volume of 4.00 mL. To investigate the effect of OPA concentration, the volume of added OPA in the total of 4.00 mL reaction mixture was lowered from 1.50 to 0.13 mL step by step. In addition, the effect of plasma dilution was investigated by using 0.50 mL of undiluted plasma taking directly from the deprotonization step and 2.5-fold diluted plasma. Since the studied florescence producing system is complex here (i.e., having signal from both background and analyte in addition to quenching matrix effect) in each examined experimental

Fig. 3. Changes in the emission spectra as the function of GSH concentration in the different amounts of OPA and plasma.

B. Hemmateenejad et al. / Talanta 79 (2009) 648–656 Table 1 The slope and intercept of calibration curves obtained at different amounts of OPA and plasma. OPA volume (mL)

1.50 1.00 0.50 0.25 0.13

Undiluted plasma

2.5-Fold diluted plasma

Slope

Intercept

Slope

Intercept

10.31 10.41 23.24 26.20 13.12

76.11 83.61 60.33 53.78 55.22

40.06 39.32 56.62 63.67 20.45

140.57 119.56 82.99 32.96 73.30

condition a calibration curve was derived and its slope and intercept were used as scoring functions for optimization. The changes in the emission spectra as the function of GSH concentration in the different amounts of OPA and plasma are shown in Fig. 3 and the slope and intercept of the corresponding calibration graphs are represented in Table 1. As it is observed from Fig. 3, when undiluted plasma is used two clear emission peaks are observed for high levels of OPA volume (i.e., 1.50 and 1.00 mL): one for GSH–OPA (420 nm) and another for plasma constituents–OPA adducts (520 nm). The peak at 520 nm is changed to a weak shoulder for OPA volume of 0.5 mL and it disappears when lower levels of OPA are used. This can be related to decreasing in the concentration of plasma constituents–OPA adduct by lowering the OPA level in the reaction mixtures. The decrease in the emission intensity at 520 nm is accompanying with increasing in the emission intensity at 420 nm due to the decreasing in the concentration of absorbing species at 420 nm (i.e., plasma constituents-OPA adduct). However, for OPA volume lower than 0.25 mL, a reduction in the signal intensity at 420 nm is observed, which can be attributed to no sufficient OPA for

653

reaction with GSH. Thus, OPA volume of 0.25 mL can be considered as the optimum value. Although the quenching effects of plasma-OPA adducts were lowered dramatically by lowering OPA volume from 1.5 to 0.25 mL, the plasma quenching effect is still present. As it is observed from right part of Fig. 3, a significant enhancement in the fluorescence emission intensity of GSH–OPA adduct was obtained by 2.5-fold dilution of plasma before addition to reaction mixture. In addition, a decrease in peak intensity is observed at 520 nm for diluted plasma. The changes in the emission intensity at 420 nm as a function of OPA volume for diluted plasma represented the similar trend as undiluted plasma and the emission intensity reached to its maximum value when OPA volume decreased from 1.5 to 0.25 mL. The slope and intercept of the calibration curves reported in Table 1 reveal the same trend as discussed in Fig. 3. The largest slope (or the best sensitivity) is observed for OPA volume of 0.25 mL. The intercept values indicate the contribution of plasma constituents (both GSH and other constituents) in the plasma fluorescence. For diluted plasma it reaches to its minimum value at OPA volume of 0.25 mL. At higher OPA levels, large intercepts are observed, which can be attributed to fluorescence of plasma–OPA adducts. According to the aforementioned results and discussion, a 0.25 mL of 0.10% (w/v) OPA and 0.5 mL of 2.5-fold diluted plasma in a total volume of 4.0 mL were considered as optimum condition for GSH assay in plasma.

3.2. PARAFAC analysis According to the aforementioned optimization results, at the suggested optimum condition the direct and indirect interferering

Fig. 4. The excitation–emission spectral profile of the mixtures of plasma constituents–OPA and GSH–OPA adducts: (a) without external GSH and (b) with 1.0 ␮M added GSH.

654

B. Hemmateenejad et al. / Talanta 79 (2009) 648–656

effects of plasma constituents are at the lowest levels. However, they still exist and thus exert their interfering effects on GSH determination in plasma. Thus, we employed PARAFAC as a multivariate calibration method to handle non-modeled interfering effects of plasma–OPA adducts. Also, PARAFAC was combined with standard addition method to overcome plasma matrix effects. For a detailed understanding of the basis of such analytical approach, the reader is referred to the literature [42–47]. However, as a short description, we expect the calibration line (i.e., PARAFAC score in the direction of added analyte concentration) to cross the Y-axis (with a statistically significant positive intercept) because there is always a basic amount of endogenous analyte (here, GSH) in the assay sample. As can be mathematically proved, the point at the intersection of the calibration curve with the horizontal axis representing, in our case, the added GSH concentrations, will be considered as the intrinsic plasma GSH level. The three-dimensional excitation–emission fluorescence spectra of plasma–OPA is shown in Fig. 4a and those of plasma added to it 1.0 ␮mol L−1 standard GSH (the first standard addition solution) is shown in Fig. 4b. Similar spectra were recorded for the other standard addition samples. The three-way data array of the excitation–emission-added GSH concentration fluorescence data were subjected to PARAFAC analysis for obtaining the relevant scores of the mentioned ways. A non-negativity constraint was exerted on the excitation and emission ways thus allowing the fitting process to work in such a way that the outcoming emission and excitation spectra follow a conventional peak shape with no negative values. No other types of constraints were found to be necessary; maybe because of the matter of PARAFAC uniqueness. A critical step in PARAFAC calibration method, similar to other factor analysis-based calibration methods, is obtaining the number of significant principal factors. For determining the proper number of components in PARAFAC, several powerful tools are available among which we used residual sum of squares, core consistency [46] and the number of iterations. A five-level repetition with random initialization strategy was specified to separately derive models based on 1–4 components. Such repetition level could ideally give opportunity to well evaluate model stability in any of the four factors. In fact, if too many components (or factors) are chosen, the number of local minima will drastically increase and as a consequence, the model becomes unstable on each run. The results are depicted in Fig. 5. All employed criteria suggest that our three-way data set is best represented by two-component PARAFAC model. We expect residual sum of squares to continually decrease upon the increase in the number of components. However, if such decrease is not statistically significant, then it will not mean that

the corresponding component(s) fall(s) among the true number of components [46]. As it is observed from Fig. 5, by moving from the first component to the second one, we observe a relatively significant decrease whereas this is not so when the next two components are considered. Core consistency is another diagnostic tool. Values obtained from this test fall in a range between zero (and maybe even negative values) and 100. If an n-component model generates core consistency values as 100 or near it, then n is very probable to be the right number of components. Values approaching zero are representative of the models the component number of which is not properly chosen. Returning to Fig. 5, we encounter a very sharp fall in the core consistency value from a two-component model to a three-component one thus indicating that there are actually two underlying factors in our fluorescence data. Models which do not converge at the same minimum mostly require an increased number of iterations to fit the data. This is probably because of the too many components selected. That is also the case when we consider the number of iterations as a determining criterion in our study. As shown in the last part of Fig. 5, a drastic scattering is clearly detectable in the values of the iteration number just after the second component. All these findings together with the visual evaluation of the three resolved modes confirm that there are only two types of chemical undergoing factors in the reaction mixture. These components can be attributed to GSH–OPA and plasma constituents–OPA adducts. Since in the experimental setup of standard addition method the concentration of all plasma constituents, except GSH was remained constant, PARAFAC was not able to discriminate between these constituents and all of them are considered as a single component, we refer to them plasma (or matrix) component. The PARAFAC scores of the resolved components for emission and excitation profiles are represented in Fig. 6. The unimodal emission profile (labeled as factor 1), representing the highest intensity at 420 nm, is definitely belonging to the GSH–OPA adduct as was also instrumentally confirmed in the phosphate buffer solution (see Fig. 1). The second emission profile does not seem to belong to only one chemical entity. Based on the fact that plasma texture also has emission properties as shown earlier in this work, the first descending part of the resolved spectrum, from wavelengths 370 nm to almost 420 nm, can be very well attributed to the plasma background emission. However, the relatively broadpeak spectrum just following that of plasma, reflects the presence of the other SH-carrying moieties that react with OPA in the same way as GSH does but presenting their emission peak intensities at wavelengths longer than 420 nm. Russel and Scaduto [35] have also reported such interfering phenomenon by presenting similar broad-peak spectra of the interfering agents. The excitation profiles resolved by the two-component PARAFAC fitting (Fig. 6) are also in harmony with the experimental findings. According to the emission profiles, the first factor can be attributed to the excitation spectrum of GSH–OPA adduct. It shows a peak at wavelengths around 340 nm, which is the same as that found for GSH–OPA adduct in the buffer solution of pH 8.0. The second excitation factor is supposed to be that of both plasma background and interfering OPA-derivatives of the plasma compounds. It exhibits a peak at 340 nm and a probable one at the higher wavelengths. However, we were not allowed to extend the excitation wavelengths higher than that of the starting wavelength of the emission profile. Russel and Scaduto have also reported that OPA-derivatives of both GSH and other potential interfering species have nearly similar excitation spectra [35]. 3.3. Application to plasma sample

Fig. 5. PF-test plots for the determination of number of factors in PARAFAC.

Once the PARAFAC scores of emission and excitation profiles were assigned to the relevant species (i.e., they were used for

B. Hemmateenejad et al. / Talanta 79 (2009) 648–656

655

Fig. 6. The resolved two-component PARAFAC scores in the directions of (a) excitation and (b) emission wavelengths. First and second scores can be attributed to GSH–OPA and plasma–OPA adducts, respectively.

qualitative analysis), the PARAFAC scores of the third way (i.e., concentration mode) was used for quantitative analysis of GSH in plasma. As it is observed from Fig. 7, the first factor is actually related to GSH while the second factor could be attributed to the other fluorescent derivative(s) generated upon reaction with OPA. As it is expected, the concentration profile of the second factor is almost constant at near zero values since there have been no change in the amount of other plasma constituents upon addition of OPA. The analytical appraisals of the resulted standard addition line are given in Table 2. There is observed a fairly linear relationship (R2 = 0.999) between the first score and the added GSH concentration. As it is shown in the lower part of Fig. 7, this calibration curve has an intercept of 444.53, which is much higher than the three times of its standard error (102.1) and therefore it can be considered significantly different from zero. Extrapolation of the calibration curve makes an intersecting point on the concentration axis with a value of 0.305 ± 0.069. This value can also be calculated by dividing intercept by slope. The standard deviation in the predicted concentration was calculated based on standard errors of the slope and intercept of the standard addition graph. This is not really the plasma intrinsic GSH concentration since we have not yet corrected this value for the dilutions plasma samples underwent. These plasma samples were 2.5 times diluted during optimization process and then 0.50 mL portions of such plasma samples were

made up to 4.0 mL as the final reaction mixture volume. Therefore, considering such successive dilutions (2.5 × 8 = 20-fold dilution), we found the plasma GSH concentration to be 6.10 ± 1.37 ␮mol L−1 . The plasma GSH concentrations reported in the literature [49–53] are in the range of 2.22–11.26 ␮mol L−1 . Thus the obtained value by the PARAFAC method complies with the literature values. To validate the accuracy of the proposed method for the assay of GSH in plasma and to obtain a confidence about the obtained GSH concentration in plasma the following procedure was employed. The series of six plasma solutions containing variable amounts of GSH standard solutions (i.e., those used for the assay of GSH in plasma) were spiked with a constant amount of standard GSH (final concentration of 1.00 ␮mol L−1 ). The three-way excitation–emission spectra of the resulting solutions were subjected to PARAFAC calibration in the same manner as described for non-spiked samples. The analytical appraisals of the resulted standard addition line (plotting the first concentration-related PARAFAC score against the concentration of added standard) are also given in Table 2. It is observed that the calibration graph of the spiked plasma has a slope very close to the non-spiked plasma solution. This explains that the amount of GSH in plasma dose not affect on the sensitivity of the PARAFAC calibration. The recovery of GSH assayed in the spiked plasma is relatively equal to the sum of that found in the non-spiked plasma and that added to it. There is observed a

Table 2 Analytical appraisals of the PARAFAC standard addition calibration lines for the analysis of GSH in plasma.

Non-spiked plasma Plasma spiked with 1.00 ␮M GSH

Slope

Intercept

R2

Found GSH

Recovery

1453.7 (±21.3) 1440.7 (±24.4)

444.53 (±102.1) 1844.0 (±130.7)

0.9991 0.9989

0.305 (±0.069) 1.280 (±0.09)

– 97.5%

656

B. Hemmateenejad et al. / Talanta 79 (2009) 648–656

method for the direct determination of GSH in plasma without need for separation. Acknowledgement Financial support of this project by the Iran National Science Foundation (INSF) is appreciated. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] Fig. 7. Plot of the concentration-related two-component PARAFAC scores against the concentration of added GSH to plasma. The filled markers denote the score of the plasma constituents.

[27] [28]

high percent of recovery for the spiked plasma (i.e., 97.5%) indicating the high accuracy of the proposed PARAFAC calibration method for the assay of GSH in plasma.

[29] [30] [31] [32] [33] [34]

4. Conclusion We developed a rapid, simple and selective method for the direct determination of GSH in plasma. It works based on PARAFAC analysis of the excitation–emission matrix fluorescence spectra of GSH upon reaction with OPA. While PARAFAC analysis could overcome the spectral interfering effects of fluorescent plasma constituents (other than GSH), its combination with standard addition method enabled us to handle the indirect interfering effect of plasma matrix. The two-component PARAFAC model resulted in pure excitation and emission fluorescence spectra of GSH–OPA and plasma constituents–OPA adducts. The first component of the third PARAFAC score represented a fair linear correlation with the concentration of added GSH (similar to that obtained in univariate standard addition method) and the second component remained unchanged as function of GSH concentration. The concentration of GSH in plasma, estimated from the regression equation of the resulted standard addition plot, was found to be 6.10 ± 1.37 ␮mol L−1 . The accuracy of the proposed method was validated by spiking standard GSH to plasma and recovering the spiked value. A recovery of 97.5% confirmed the ability of the employed

[35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53]

A. Meister, Biochim. Biophys. Acta 1271 (1995) 35. D.J. Reed, M.K. Savage, Biochim. Biophys. Acta 1271 (1995) 43. A. Meister, Science 220 (1983) 472. A.J.L. Cooper, B.S. Kristal, Biol. Chem. 378 (1997) 793. A. Meister, Cancer Res. 54 (1994) 1969. A. Jain, J. Martensson, E. Stole, P.A. Auld, A. Meister, Proc. Natl. Acad. Sci. U.S.A. 88 (1991) 1913. S. Luprecio, S. Tamir, S.R. Tannebaum, Free Radic. Biol. Med. 21 (1996) 513. M. Poot, H. Teubert, P.S. Rabinovitch, T.J. Kavanagh, J. Cell Physiol. 163 (1995) 555. D.J. van den Dobbelsteen, C.S.I. Nobel, J. Schlegel, I.A. Cotgreave, S. Orrenius, A.F. Slater, J. Biol. Chem. 271 (1996) 15420. L. Ghibelli, C. Fanelli, G. Rotilio, E. Lafa via, S. Coppolia, C. Colussi, FASEB J. 12 (1998) 479. A.G. Hall, Eur. J. Clin. Invest. 29 (1999) 238. G. Bellomo, F. Mirabelli, D. Di Monte, P. Richelmi, H. Thor, C. Orrenius, S. Orrenius, Biochem. Pharmacol. 36 (1987) 1313. Y.C. Chai, S.S. Ashraf, K. Rokutan, R.B. Johnston Jr., J.A. Thomas, Arch. Biochem. Biophys. 310 (1994) 273. E. Camera, M. Picardo, J. Chromatogr. B 781 (2002) 181. K.J. Lenton, H. Therriault, J.R. Wagner, Anal. Biochem. 274 (1999) 125. B.R. Shivakumar, S.V. Kolluri, V. Ravindranath, Mol. Cell. Biochem. 111 (1992) 125. A. Patore, G. Federici, E. Bertini, F. Piemonte, Clin. Chim. Acta 333 (2003) 19. B.J. Mills, J.P.J. Richie, C.A. Lang, Anal. Biochem. 184 (1990) 263. M. Floreani, M. Petrone, P. Debetto, P. Palatini, Free Radic. Res. 26 (1997) 449. M. Koller, H. Eckert, Anal. Chim. Acta 352 (1997) 31. D. Tsikas, J. Sandmann, D. Holzberg, P. Pantazis, M. Raida, J.C. Frolich, Anal. Biochem. 273 (1999) 32. I. Molnar-Perl, J. Chromatogr. A 913 (2001) 283. V.H. Cohn, Lyle, Anal. Biochem. 14 (1966) 434. P.J. Hissin, R. Hilf, Anal. Biochem. 74 (1976) 214. A.R. Ivanov, I.V. Nazimov, L. Baratova, A.P. Lobazov, G.B. Popkovich, J. Chromatogr. A 913 (2001) 315. M.A. Raggi, R. Mandrioli, C. Sabbioni, F. Mongiello, M. Marini, S. Fanali, J. Microcolumn Sep. 10 (1998) 503. K. Imai, S. Uzu, T. Toyo’oka, J. Pharm. Biomed. Anal. 7 (1989) 1395. I.K. Abukhalaf, N.A. Silvestrov, J.M. Menter, J. Pharm. Biomed. Anal. 28 (2002) 637. B. Tang, Y. Xing, P. Li, N. Zhang, F. Yu, G. Yang, J. Am. Chem. Soc. 129 (2007) 11666. M. Marcos, P.J. Chmielewski, Org. Lett. 10 (2008) 837. R. Paroni, E. De Vecchi, G. Cighetti, et al., Clin. Chem. 41 (1995) 448. P.C. Jocelin, A. Kamminga, Anal. Biochem. 37 (1970) 417. E. Beutler, C. West, Anal. Biochem. 81 (1977) 458. R. Brigelius, C. Muckel, T.P.M. Akerboom, H. Sies, Biochem. Pharmacol. 32 (1983) 2529. C. Russel, J.R. Scaduto, Anal. Biochem. 174 (1988) 265. M. Hasani, M. Moloudi, F. Emami, Anal. Biochem. 370 (2007) 68. A.A. Ensafi, T. Khayamian, R. Tabaraki, Talanta 71 (2007) 2021. B. Hemmateenejad, A. Abbaspour, H. Maghami, R. Miri, M.R. Panjehshahin, Anal. Chim. Acta 575 (2006) 290. Z. Rezaei, B. Hemmateenejad, S. Khabnadideh, M. Gorgin, Talanta 65 (2005) 21. M. Shamsipur, J. Tashkhourian, B. Hemmateenejad, H. Sharghi, Talanta 64 (2004) 590. K. Kargosha, A.H.M. Sarrafi, J. Pharm. Biomed. Anal. 26 (2001) 273. ˜ R.J. G.M. Escandar, A.C. Olivieri, N.M. Faber, H.C. Goicoechea, A.M. de la Pena, Poppi, Trends Anal. Chem. 26 (2007) 752. H. Abdollahi, A. Safavi, S. Zeinali, Chemometr. Intell. Lab. Syst. 94 (2008) 112. K. Vatsavai, H.C. Goicoechea, A.D. Campiglia, Anal. Biochem. 376 (2008) 213. ˜ F. Canada-Ca ˜ ˜ A. Espinosa-Mansilla, A.M. de la Pena, nada, D.G. Gómez, Anal. Biochem. 347 (2005) 275. Rasmus Bro, Chemometr. Intell. Lab. Syst. 38 (1997) 149. C.M. Anderson, R. Bro, J. Chemometr. 17 (2003) 200. A. Brent, T. Neuschwander, F. Jozeph Roll, Anal. Biochem. 179 (1989) 236. F. Michelet, R. Gueguen, P. Leroy, et al., Clin. Chem. 41 (1995) 1509. J.D. Adams, J.N. Johannesen, J.P. Bacon, Clin. Chem. 33 (1987) 1675. J.D. Adams, E. Khaw, G. Jaresko, et al., Res. Commun. Subst. Abuse 11 (1990) 153. M. Kretzschmar, D. Muller, J. Hubscher, et al., Int. J. Sports Med. 12 (1991) 218. M.A. Mansoor, A.M. Svardal, P.M. Ueland, Anal. Biochem. 200 (1992) 218.