GSH, GSSG and trifluoroacetic acid (TFA) were purchased from Sigma-Aldrich (St. Louis, MO, USA). HPLC-grade methanol (>99.9%; v/v) was obtained from Merck KGaA (Darmstadt, Germany). Other chemicals were purchased from Sigma-Aldrich unless otherwise stated. Stock standard solutions of the thiols (1 mg.ml⁻¹) were prepared with ACS water (Sigma-Aldrich) and stored at −20°C in the dark. Working standard solutions were prepared daily by diluting the stock solutions. All solutions were filtered through 0.45-μm nylon filter discs (Millipore, Billerica, MA, USA) prior to HPLC analysis. The pH value was measured using WTW inoLab Level 3 with terminal Level 3 (WTW GmbH, Weilheim, Germany).

The HPLC-ED system consists of two chromatographic pumps (Model 582; ESA, Inc., Chelmsford, MA, USA; working range 0.001–9.999 ml/min), a chromatographic column with reverse phase Zorbax eclipse AAA C18 (Agilent Technologies, Inc., Santa Clara, CA, USA; 150x4.6 mm; 3.5-μm particles) and a twelve-channel CoulArray electrochemical detector (Model 5600A; ESA, Inc.). The detector consists of three flow analytical chambers (Model 6210; ESA, Inc.). Each chamber contains four analytical cells and one analytical cell contains two referent (hydrogen-palladium), as well as two counters and porous graphite working electrodes. The ED is situated in the thermostated control module. A 20 μl sample was injected using an autosampler (Model 542; ESA, Inc.), which has thermostated space for the column. The column was termostated at 35°C. Other conditions were optimised and are described later.

Recovery of GSH and GSSG were evaluated with homogenates spiked with standards according to Causon (50). Prior to extraction, 100 μl GSH and GSSG was added to the blood serum homogenate. Homogenates were blindly assayed and the concentration of GSH and GSSG was derived from the calibration curves. The spiking of GSH and GSSG was determined as a standard measured in the absence of real sample. Accuracy was evaluated by comparing the estimated concentration with the known concentrations of both thiols.

Blood samples were obtained from 116 children hospitalised at the Department of Paediatric Haematology and Oncology (Faculty Hospital Motol, Prague, Czech Republic) with newly diagnosed solid tumours of neuroblastoma (n=27), nephroblastoma (n=8), anaplastic ependymoma (n=4), Ewing’s sarcoma (n=9), germ cell tumour (n=4), osteosarcoma (n=16), tumour of the genital tract (n=6), Hodgkin’s lymphoma (n=16), lymphadenopathy (n=3), medulloblastoma (n=15), rhabdomyosarcoma (n=4) and retinoblastoma (n=4). The study was approved by the ethics committee of Faculty Hospital Motol, Prague, Czech Republic. Written informed patient consent was obtained from the patients. Subjects ranged between 1 and 10 years of age. The blood samples were collected before chemotherapy and radiotherapy. Serum was separated by centrifugation at 4,000 × g for 10 min (Model 5402; Eppendorf AG, Hamburg, Germany), and the samples were stored at −80°C until assayed. When required, the denatured samples were centrifuged at 15,000 x g at 4°C for 30 min (Model 5402; Eppendorf AG) and directly analysed using an optimised HPLC-ED method.

Data were processed using Microsoft Excel (USA). Results are expressed as mean ± standard deviation (SD) unless otherwise stated. The detection limits [3 signal/noise (S/N)] were calculated according to Long and Winefordner (51), while N was expressed as the SD of noise determined in the signal domain unless otherwise noted.

Primarily, it was necessary to optimise the separation and subsequent ED in order to achieve the required accuracy and sensitivity for the determination of GSH and GSSG in real blood serum samples. Therefore, we focused on studying the influence of flow rate, concentration of components of the mobile phase, elution and applied potential of the working electrodes on GSH and GSSG signals.

The mobile phase flow rate is an important parameter influencing the electrochemical response of the detector. When using a chromatographic column Zorbax Eclipse AAA, optimum mobile phase flow rate was 1 ml/min at pressures of 130 bars. Additionally, we identified that if the flow >1 ml/min, the responses of GSH and GSSG decreased by >10%. This is probably caused by reducing the time-concentration of the analyte on the electrode surface. Even with a lower flow rate, a decreased signal occurred compared with the maximum, but the total peak area remained the same with a tolerance of 7%. Although a lower flow rate may not be significantly affected by resolution, it may extend the period of separation, which is critical for analysing a large number of clinical samples. Therefore, we decided to use 1 ml/min as the optimum flow rate of the mobile phase.

Achieving an optimal resolution is crucial for simultaneous separation of analytes. In order to separate all determined substances in the system with reversed-phase, a gradient with the increasing content of organic solvent is required. Since the electrochemical determination of substances contained in the sample requires the presence of an electrolyte, we examined the effect of the organic solvent (methanol) on the electrochemical response of analytes. We identified that 15% content of methanol in the mobile phase, which is the polar component of the mobile phase composed also from 80 mM TFA, lead to more than 50% decrease in GSH signal. A marked decline of GSSG signal was also observed. The best ratio of 80 mM TFA and methanol in the mobile phase was 99:1 (v/v).

If GSH and GSSG were separated by isocratic elution where the ratio of TFA and methanol was 99:1 (v/v), it would be the most sensitive, but the retention times of the separated substances would be too high. A significant tailing of peaks was observed during the elution of compounds with higher retention under these conditions. Therefore, we optimised the increasing content of methanol with respect to the sensitivity of the analysis. Based on the optimisation steps, the mobile phase, which consisted of (A) 80 mM TFA and (B) 100% methanol, was used for separation and detection of GSH and GSSG. Compounds were eluted by following an increasing linear gradient: 0–1 min (3% B), 1–2 min (10% B), 2–5 min (30% B) and 15–16 min (98% B). Flow rate of the mobile phase was 1 ml/min, and the time of one analysis inducing column regeneration was 20 min.

Sensitivity of the electrochemical detector may be more influenced by factors including the type of electrolyte in the mobile phase, concentration, pH and, in particular, applied potential. TFA was used as an ion-pair reagent, which provides the best separation conditions in the parameters mentioned above, and at a concentration of 80 mM it is also an extremely suitable electrolyte for the detection of thiols. We further studied the effect of the applied potential on the working electrode set separately for GSH and GSSG, which were designed for hydrodynamic voltammogram (HDV). Tested potentials were 100, 200, 300, 400, 500, 600, 700, 800, 900 and 1,000 mV. The responses detected at 100 mV were negligible; however, when the potential reached 300 mV, detectable signals for GSH and GSSG were observed. While the GSH signal markedly increased from 600 mV, the GSSG signal markedly increased from 700 mV. This is probably due to the requirement for greater power for partial dissociation of the -S-S- group on the surface of the working electrode, in comparison to the relatively easily accessible -SH moiety of GSH. We observed the highest signals for both compounds when a potential of 900–1,000 mV was applied, which is evident from the HDVs showed in Fig. 2A. Based on the HDV results we were able to evaluate that the best glutathione detection was achieved when a potential of 900 mV was applied to the working electrodes.

After identifying the optimal separation and detection conditions, the calibration curves for GSH and GSSG were measured within the concentration range of 0.2–100 μM. Overlay of HPL chromatograms is shown in Fig. 2B, and the calibration curves are shown in Fig. 2C. The obtained dependences were strictly linear with R²=0.9997 for GSH and R²=0.9936 for GSSG. Detection limits (3 S/N) were estimated with nanomolar subunits for both substances of interest.

Prior to chromatographic analysis, precipitation of proteins with TFA to avoid excessive clogging of filters and precolumns, which protect the separation column from contaminations, was required. The proteins may interfere with detected substances and the obtained chromatograms may be extremely difficult to analyse. The denatured sample was than centrifuged and the resulting supernatant was directly injected to the chromatographic column. To ensure the lowest possible loss of target compounds during sample preparation it was necessary to examine several factors of a sample treatment, which could affect the overall recovery of GSH and GSSG.

Given that the formation of complexes may be faster under certain conditions (pH and ionic strength), we decided to investigate the possibility of GSH complex formation in the solution used for isolation. The formation of the complex was determined via a decrease in the GSH peak. Primarily, we examined the effect of molar concentrations of phosphate buffer (0.1, 0.5, 1, 5, 10 and 20 mM; pH 7.5) on the GSH (50 μg/ml) signal. These samples were left following preparation at room temperature, and were analysed by HPLC at time 0, 130 (2.2 h) and 260 min (4.3 h). Based on the results obtained, higher concentrations of buffer caused a decreasing GSH signal, i.e. concentration; thus, 20 mM phosphate buffer caused the highest decrease of glutathione concentration. It is clear that the greatest stability of GSH was observed in samples prepared in the presence of low concentrations of phosphate buffer.

Specifically, the lowest loss of glutathione occurred at the applied concentrations of 0.1–1 mM (Fig. 3A). These results clearly demonstrated that lower concentrations of phosphate buffer contribute to the stability of the sample. Therefore, for further experiments we used 1 mM phosphate buffer (pH 7.5).

To determine the extent of oxidative stress by glutathione it is necessary to know the ratio of GSH:GSSG. Therefore, we were aimed to determine whether TFA, which is normally added to the sample due to deproteination, could have an effect on GSSG level. We also studied the effect of adding the reducing agent tris(2-carboxylethyl)phosphine (TCEP), which may markedly influence the ratio of GSH:GSSG. Studies on TFA and TCEP were conducted in buffer and blood sera, and all variants were prepared with the same concentration of 50 μg/ml GSH and 5 μg/ml GSSG. Samples were prepared in the presence of (i) 1 mM phosphate buffer (pH 7.5), (ii) 1 mM phosphate buffer (pH 7.5) with 5% TFA (v/v), and (iii) 1 mM phosphate buffer (pH 7.5) with 1 mM TCEP. To be able to assess the influence of the matrix, samples of blood serum were prepared in the same way. All samples were vortexed for 1 min and immediately analysed by HPLC following preparation. The GSH:GSSG ratio was determined, where the ratio of 10 was taken as a control. In the case of using 5% TFA, ±7% change from control was determined in variants of buffer and serum (Fig. 3B). The results reveal that TFA did not affect the ratio of GSH:GSSG. However, following the addition of TCEP, there was a significant increase in the ratio to 38 and 48 in the buffer and blood serum, respectively. TCEP reduced the majority of GSSG to GSH, which was the reason for the significant increase of the GSH:GSSG ratio. In the case of blood serum, the ratio was even higher compared with buffer. This phenomenon may be explained by the involvement of the biological matrix in a non-specific reaction of the complexes or the presence of certain concentrations of glutathione bound to the matrix constituents. These results clearly indicate that TCEP reduces GSSG back to GSH, which could be used to determine the total amount of glutathione.

Recovery estimation for sample preparation and analysis for a sample of blood serum using an optimised separation method was conducted by adding 10 μg/ ml GSH and 10 μg/ml GSSG prior to precipitation with 5% TFA and subsequent centrifugation. A sample with a GSH:GSSG ratio of 2.8 was used for determining recovery. The resulting recoveries are indicated in Table I. A recovery estimation of 83 and 89% for GSH and GSSG, respectively, clearly follows from the results previously obtained. GSH recovery can be associated with the imperfect protection of free-SH groups of glutathione, which can interact with the remains of biological matrices, and thus reduce the total concentration of free GSH during the preparation of the samples.

The antioxidant function of GSH is primarily due to its involvement in enzymatic pathways that cells have developed against ROS. The most important pathway involves glutathione peroxidase (GPx) and glutathione reductase (GR). GPx catalyses the reduction of hydrogen peroxide, which is produced by superoxide dismutase (SOD) through the dismutation of superoxide anions or organic hydroperoxides. GSH and GSH-dependent enzymes act in cooperation to scavenge ROS and/or neutralise their toxic oxidising effect. These systems act at the same time and in cooperation to protect the human body from ROS. Under oxidative stress conditions, GSH is oxidised to GSSG; thus, the GSH:GSSG ratio is altered.