Validation of a simplified procedure for convenient and rapid quantification of reduced and oxidized glutathione in human plasma by liquid chromatography tandem mass spectrometry analysis

Abstract Endogenous glutathione (GSH) and glutathione disulfide (GSSG) status is highly sensitive to oxidative conditions and have broad application as a surrogate indicator of redox status in vivo. Established methods for GSH and GSSG quantification in whole blood display limited utility in human plasma, where GSH and GSSG levels are ~3–4 orders of magnitude below those observed in whole blood. This study presents simplified sample processing and analytical LC–MS/MS approaches exhibiting the sensitivity and accuracy required to measure GSH and GSSG concentrations in human plasma samples, which after 5‐fold dilution to suppress matrix interferences range from 200 to 500 nm (GSH) and 5–30 nm (GSSG). The utility of the methods reported herein is demonstrated by assay performance and validation parameters which indicate good sensitivity [lower limits of quantitation of 4.99 nm (GSH) and 3.65 nm (GSSG), and high assay precision (intra‐assay CVs 3.6 and 1.9%, and inter‐assay CVs of 7.0 and 2.8% for GSH and GSSG, respectively). These methods also exhibited exceptional recovery of analyte‐spiked plasma samples (98.0 ± 7.64% for GSH and 98.5 ± 12.7% for GSSG). Good sample stability at −80°C was evident for GSH for up to 55 weeks and GSSG for up to 46 weeks, with average CVs <15 and <10%, respectively.

Although the concentration of, and the ratio between, GSH and GSSG within various cells and tissues have demonstrated relationships between oxidative status and specific disease states (Ballatori et al., 2009;Fernández-Checa et al., 1997;Giustarini et al., 2011;Lu, 2013), the use of these indicators for the evaluation of redox status in human plasma has previously been limited by a wide range of factors (Claeson, Gouveia-Figueira, Stenlund, & Johansson, 2019;Forgacsova et al., 2019). In particular, the relatively low concentration of GSH and GSSG in plasma (i.e. relative to other biological sample types) (Jones et al., 1998), as well as the poor sensitivity of traditional assay methods (Camera & Picardo, 2002) has presented a major challenge in terms of accurately and precisely quantifying GSH and GSSG in plasma at physiological concentrations in clinical research. In addition, methods such as HPLC-UV and HPLCfluorescent spectroscopy that are currently utilized to detect GSH and GSSG at lower concentration ranges typically require laborious processing procedures and technically challenging preparation techniques (e.g. chemical derivatization) in order to ensure reliable measurements (Jones et al., 1998). Finally, many methods used to quantify GSH and GSSG are limited to applications utilizing specific sample types [i.e. whole blood or red blood cell (RBC) lysate] that are incompatible with plasma samples (Giustarini et al., 2011;Giustarini, Dalle-Donne, Milzani, Fanti, & Rossi, 2013;Squellerio et al., 2012).
In addition to the methodological limitations noted above, numerous physico-chemical factors affect GSH and GSSG stability, including sample pH as well as the composition of buffers and extraction reagents, which prevents the acquisition of reliable estimates of GSH and GSSG in many biological samples (Nishiyama & Kuninori, 1992), including human plasma (Jones et al., 1998). GSH is also susceptible to artifactual oxidations, which can markedly change both GSH and GSSG estimates ex vivo, thereby leading to misrepresentations of endogenous redox states (Giustarini et al., 2011;Jones et al., 1998;Nishiyama & Kuninori, 1992). Moreover, GSH and GSSG are susceptible to degradation and chemical modifications by proteolytic and phase II metabolic enzymes (e.g. γ-glutamyltranspeptidases and glutathione-s-transferases) and glutathione reductases that can alter GSH and GSSG concentrations within samples during processing and storage (Jones et al., 1998;Lu, 2013). Notably, although certain protective groups have been utilized to stabilize GSH and GSSG levels in samples through the formation of thiolmasked adducts, these "protected" derivatives are also sensitive to subtle fluctuations in pH and temperature and therefore do not unequivocally ensure accurate detection and quantification of GSH and GSSG (Giustarini et al., 2013;Nishiyama & Kuninori, 1992;Roosild et al., 2010).
In spite of the many factors that confound the assessment of GSH and GSSG concentrations in plasma, several groups have published methods to measure glutathione levels in animal as well as human plasma (Claeson et al., 2019;Forgacsova et al., 2019;Jones et al., 1998). However, technical limitations, particularly with regard to human plasma analysis of GSH and GSSG, limit their application in clinical research. In the current study, we describe a simplified plasma processing procedure along with convenient LC-MS/MS methods that provides a simple, convenient, accurate and precise method to quantify GSH and GSSG in human plasma sample stored for up to 55 weeks and 46 weeks, respectively. These attributes make this assay well suited for clinical research in which long-term storage is often a necessity, and subtle changes in GSH or GSSG levels in plasma or serum are indicative of changes in health status, as has been observed, for example, in disease states such as diabetes (Costagliola et al., 1990), cystic fibrosis (Roum, Buhl, McElvaney, Borok, & Crystal, 2017) and HIV (Borges-Santos, Moreto, Pereira, Ming-Yu, & Burini, 2012), as well as in aging (Jones, Mody, Carlson, Lynn, & Sternberg, 2002 (hereafter referred to as GSSG-IS) for GSSG were prepared in 0.1 M PB at concentrations of 100 and 10 μM, respectively, and then stored in single-use aliquots at −80 C until use. GSH-NEM, GSH-NEM-IS, GSSG and GSSG-IS working standards and calibration standards were prepared daily in assay buffer (0.1 M PB with 5 mM NEM) from stock solutions as described above.

| Sample collection, plasma processing and storage
Human venous blood samples were drawn into 10 ml BD Vacutainer ® collection tubes (366,401, BD Life Sciences) containing K 2 -EDTA (lavender cap) by a certified clinician. Samples were processed immediately following collection as follows: vacutainers containing human blood samples were centrifuged at 1200 rcf for 12 min at 4 C immediately following collection to separate plasma. After centrifugation vacutainers were placed on ice and plasma was collected and trans-  Analytical recovery, or percentage recovery, of GSH and GSSG was evaluated by calculating the recovery of GSH or GSSG spiked into individual or pooled plasma samples as follows:

| Analyte stability
The stability of GSH-NEM and GSSG in NEM-treated plasma samples following storage at −80 C was examined at multiple time points ranging from 1 week to over 1 year after collection. Specifically, at 1, 12, 46 and 55 weeks post-collection, frozen NEM-treated plasma sample aliquots maintained at −80 C were allowed to thaw and equilibrate to RT. Each sample (100 μl) was transferred into a new microcentrifuge tube then processed as described in Section 2.3.
GSH and GSSG levels in processed plasma extracts where determined following the analytical procedures previously described in Section 2.4.

| Method validation
Data for assay specificity are presented in Table 1 Table 2. Intra-assay variation, expressed as the coefficient of variation (CV), was 3.7% for GSH-NEM and 1.9% for GSSG. The CV(%) for reproducibility (inter-assay variation) was 7% for GSH-NEM and 2.8% for GSSG. All values were within the ±15 CV(%) limit defined by the US Food and Drug Administration (2018).
Recovery of GSH and GSSG was determined by comparing the measured concentration of each analyte in plasma before and after spiking with a known, physiologically relevant quantity of GSH or GSSG (Table 3). The average GSH-NEM spike recovery was 98.0 ± 7.64%, whereas the average spike recovery of GSSG was 98.5 ± 12.7%.

| Matrix effects
There was a significant matrix effect of the plasma sample extract on the ionization intensity of GSH, but not GSSG (Figure 3). Regression analysis (i.e. slope of the line) indicated that the ionization intensity (AUC) of GSH-NEM decreased by 5.6% when evaluated in plasma matrix compared with buffer ( Figure 3a). This decline was statistically significant (ANCOVA, P = 0.02). In contrast, GSSG ionization intensity was not affected by the plasma matrix. The regression slope of GSSG in plasma was reduced by a nonsignificant (ANCOVA, P = 0.432) 2.1% compared with the regression slope of GSSG in buffer.

| DISCUSSION
Early assessments of GSH and GSSG used techniques that did not address a number of factors now known to introduce artifactual interference (Rossi et al., 2006). As such, the validity of data derived from T A B L E 2 Intra-and inter-assay variation of GSH and GSSG in a pooled NEM plasma sample analyzed 13 times within one day (intra-assay) or five times daily for 3 days (inter-assay)

Intra-assay (n = 13)
Inter-assay (n = 15) Average (μM) Standard deviation (μM) CV (%) Average (μM) Standard deviation (μM) CV (%) Indicates final concentration in plasma. b For GSH (free thiol), n = 3 independent pooled plasma samples; for GSSG, n = 5 independent samples.  these assays is open to question. However, more recent developments concerning sample collection and processing as well as storage and handling techniques have dramatically improved the reliability of GSH and GSSG measurements, particularly as an indicator of oxidative stress in various types of biological samples (Giustarini et al., 2011;Rossi et al., 2006). To date, ongoing research in several laboratories has established multiple strategies for GSH and GSSG determination in human whole blood and isolated RBCs that are effective (Giustarini et al., 2013;Rossi et al., 2006;Squellerio et al., 2012); however, these methods involve intensive collection and processing procedures and require proficiency in difficult/complex techniques to ensure accurate measurements. Furthermore, many of the methods that perform well with human whole-blood samples and RBC lysates do not translate well for evaluation of GSH and GSSG in human plasma samples (Giustarini et al., 2013). In the current report, we describe a new approach specifically developed to address a wide range of problems that have historically contributed to erroneous estimates of GSH and GSSG in human plasma. In this specialized procedure a novel LC-MS/MS method was developed to examine endogenous concentrations of the two major glutathione redox forms in clinical plasma samples acquired and handled using simplified collection and processing procedures (see Materials and Methods).
The methods described in this report are straightforward, can be executed in comparatively short periods of time and moreover provide accurate estimates of endogenous GSH and GSSG levels in human plasma samples that correspond to previously published values derived from methodologies that are considerably more challenging technically than the method presented herein (Claeson et al., 2019;Jones et al., 1998;Jones & Liang, 2009). As such, this technique can be applied in array settings that have not previously been available for human clinical research.
Thiol moieties are highly unstable. It is therefore unsurprising that GSH and GSSG readily undergo biochemical changes both in vivo and ex vivo. To address the potential influence of ex vivo thiol oxidation and reactivity on measurements of GSH and GSSG, we enacted three specific steps. First, we selected NEM as a protective thiolmasking reagent to prevent artifactual GSH oxidation. Indeed, for whole-blood glutathione analyses, NEM is typically added directly to the sample at the time of collection and is often added to collection tubes before drawing blood samples (Giustarini et al., 2013). However, NEM exhibits significant hemolytic activity (Kuypers et al., 1996). Therefore, in order to avoid potential contamination of plasma with RBC-derived GSH and GSSG, reported to be orders of magnitude higher in RBCs than in plasma (Giustarini et al., 2013), the separation of plasma from whole blood samples was performed in the absence of NEM at refrigeration temperatures (i.e. 2-4 C) over a brief period of time (12 min). Once collected, plasma supernatant was transferred directly into tubes prefilled with NEM. Second, as an additional measure to prevent ex vivo changes in glutathione oxidation states, we collected blood samples in vacutainers containing EDTA, a well-established chelating agent (Cotton, 2003). This was done to reduce the potential for transition-metal catalyzed GSH oxidation which is known to occur in whole blood (Squellerio et al., 2012).
Finally, because GSH displays high instability at room temperature, particularly in the absence of protective reagent or preservation buffers (Claeson et al., 2019), we processed all samples on ice or via refrigerated centrifugation (4 C), as soon as the blood draw was completed. The combined effect of each of these steps was to maintain the integrity of plasma GSH and GSSG levels, as evidenced in particular by the extremely low levels of plasma GSSG (i.e. < 100 nM) measured via this method that are as low as or lower than recently reported values (Bettermann et al., 2018;Claeson et al., 2019).
In addition to thiol reactivity, we identified two other key issues that were critical in establishing this method of GSH and GSSG analysis in human plasma. The first was buffer pH; previous research from other laboratories provides evidence that alkaline conditions may facilitate hydrolysis of GSH-NEM adducts at multiple positions, potentially generating products with different mass fragments and ionization states that could confound interpretations of LC-MS/MS analyses (Nishiyama & Kuninori, 1992;Roosild et al., 2010). Evidence suggesting the potential influence this may have on our measurements of GSH and GSSG includes observations that detection of GSH-NEM adduct was exclusively achieved when conditions were kept slightly acidic to acidic during sample preparation (data not published). We also found that ionization of GSH and GSSG was sensitive to reagents and procedures used in sample preparation and processing. For example, changing the extraction agent from TCA to MPA perturbed ionization of analytes and limited detection to a range that was insufficient for quantification of GSH and GSSG in plasma samples, even with minimal sample dilution. In addition, we found that adjusting the concentration of TCA from 10% w/v to 5% w/v: (a) exacerbated chromatography problems as a consequence of incomplete protein precipitation; and (b) significantly reduced GSH-NEM and GSSG ionization during LC-MS/MS analyses. However, the use of 10% TCA optimized both protein precipitation and GSH-NEM adduct formation and, moreover, had negligible effects on analyte ionization properties as evidenced by a minimal (i.e. 5.6%) and nonexistent matrix effect of the plasma extract on GSH-NEM and GSSG, respectively.
A number of recent publications have detailed approaches that simultaneously measure GSH and GSSG (Bondada et al., 2016;Guan, Hoffman, Dwivedi, & Matthees, 2003;Harwood, Kettle, Brennan, & Winterbourn, 2009). Although these procedures may be applicable to measuring GSH and GSSG in many types of samples, they are inadequate for the simultaneous measurement of GSH and GSSG in human plasma for several reasons. In particular, many of these methods lack the sensitivity necessary to measure GSH and GSSG in plasma, which requires LLOQs in the nanomolar to subnanomolar range. Indeed, after failing to achieve appropriate sensitivity (i.e. LLOD and LLOQ) using previously published methods that were designed for the simultaneous detection of GSH and GSSG, we adjusted our strategy to employ two separate protocols, one to measure each redox form of glutathione. By separating different analytes on separate LC columns with different mobile phases we were able to detect GSH and GSSG as distinct, well-defined, integrated peaks on individual chromatograms (Figure 1), improving sensitivity to meet the necessary range for clinical plasma analyses. Although this approach required two distinct preparations and two distinct LC-MS/MS analytical conditions in order to measure GSH and GSSG in each sample, the superior resolution of the distinct procedures markedly improved GSH and GSSG detection sensitivity, with assay LLODs in the nanomolar and subnanomolar range, respectively. In addition, performing two different procedures did not require an excessive investment of time as sample processing and preparation could be performed rapidly (about 2-3 h to prepare 20-30 samples) and the run-time for each injection was completed within 20 min for GSH and in <10 min for GSSG. Note that for GSH-NEM the run time (20 min) was considerably longer than the retention time (4.2 min).
This run time was selected empirically as a precaution to ensure removal of nontarget substances from the LC-column matrix that might otherwise interfere with or obscure the precise detection and quantification of GSH-NEM in subsequent samples. Indeed, we did not observe any sample carryover or contamination throughout the duration of this study, and attribute this in part to the extended GSH-NEM run time.
A recent study by Claeson et al. described an ultra-performance-LC-MS/MS (UPLC-ESI-MS/MS) approach for measuring GSH and GSSG in human plasma (Claeson et al., 2019). Although the method is well suited for investigating the impact of preparation and storage conditions on reduced and oxidized levels of glutathione with a reported linear range between 0.1 and 10 μM for both GSH and GSSG, this sensitivity is not adequate for identifying subtle changes in clinical treatment paradigms. In addition, the method does not involve precautionary measures to protect samples from autooxidation and the authors report notable shifts in analyte detection after 3 months of storage at −80 C. The approach detailed in this report includes simple steps to maintain analyte stability, which remains apparent over 1 year of storage (as shown in Table 3), and allows detection in a linear range from 2 to 500 nM for GSSG and from 16 nM to 2 μM for GSH.

| CONCLUSION
In conclusion, the processes and procedures we have detailed in the current work provide a highly sensitive, yet simple and rapid method for evaluation of GSH and GSSG in human plasma samples with a wide range of advantages compared with previous approaches performed alone. The detection range and stability we established for these procedures indicate that this method can identify minimal variations in plasma GSH and GSSG levels that could not be discriminated via previous methods.