Noninvasive in vivo magnetic resonance measures of glutathione synthesis in human and rat liver as an oxidative stress biomarker

Oxidative stress (OS) plays a central role in the progression of liver disease and in damage to liver by toxic xenobiotics. We have developed methods for noninvasive assessment of hepatic OS defenses by measuring flux through the glutathione (GSH) synthesis pathway. 13C-labeled GSH is endogenously produced and detected by in vivo magnetic resonance after administration of [2-13C]-glycine. We report on a successful first-ever human demonstration of this approach as well as preclinical studies demonstrating perturbed GSH metabolism in models of acute and chronic OS. Human studies employed oral administration of [2-13C]-glycine and 13C spectroscopy on a 3T clinical magnetic resonance (MR) imaging scanner and demonstrated detection and quantification of endogenously produced 13C-GSH after labeled glycine ingestion. Plasma analysis demonstrated that glycine 13C fractional enrichment achieved steady state during the 6-hour ingestion period. Mean rate of synthesis of hepatic 13C-labeled GSH was 0.32 ± 0.18 mmole/kg/hour. Preclinical models of acute OS and nonalcoholic steatohepatitis (NASH) comprised CCl4-treated and high-fat, high-carbohydrate diet-fed Sprague-Dawley rats, respectively, using intravenous administration of [2-13C]-glycine and observation of 13C-label metabolism on a 7T preclinical MR system. Preclinical studies demonstrated a 54% elevation of GSH content and a 31% increase in flux through the GSH synthesis pathway at 12 hours after acute insult caused by CCl4 administration, as well as a 23% decrease in GSH content and evidence of early steatohepatitis in the model of NASH. Conclusion: Our data demonstrate in vivo 13C-labeling and detection of GSH as a biomarker of tissue OS defenses, detecting chronic and acute OS insults. The methods are applicable to clinical research studies of hepatic OS in disease states over time as well as monitoring effects of therapeutic interventions.

averages), and the discrepancy between requested and achieved tip angle calculated from the variation in signal intensity between the four spectra. This calibration procedure was designed to enable correction of differences in achieved 13 C tip angle between study volunteers resultant from differences in coil loading. However, in practice the procedure demonstrated that all three volunteers and the calibration phantom had the same 13 C RF power requirements, reflecting the similarity in coil loading between scans.
Consistency of 1 H decoupler power was also optimised for study scans. Tests performed during coil development demonstrated consistent decoupler efficacy over the active volume of the 13 C coil above a threshold power value, and we have successfully employed acquisition of 1 H-decouped 13 C spectra in previous studies (1)(2)(3). In coil development studies we observed that automated scanner 1 H power calibration preparation steps performed prior to scan execution have the potential to introduce a variation in 1 H RF power output when scans were acquired using our 13 C/ 1 H surface coil. We attributed this to the 1 H power calibration procedure being optimised for homogeneous B 1 fields generated by the scanner's body coil rather than the inhomogeneous B 1 field of a surface coil. Thus automated 1 H power optimisation preparation steps were disabled for this study, and consistency of 1 H power output monitored via the scanner 1 H RF amplifier power meter (standard deviation of decoupler power = ± 1.7%). Efficacy of 1 H decoupling over the active volume of the 13 C coil was confirmed in phantom studies employing 1 Hdecoupled 13 C 1D chemical shift imaging.
Prior to acquisition of 13 C spectra, homogeneity of the scanner's B 0 magnetic field was optimised over a 16 x 16 x 12 cm 3 volume of interest located over the subject's liver using a B 0 -mapping method implemented by the scanner manufacturer, based on the method described by Schär et al (4), which calculates optimal scanner shims from water proton frequency measurements acquired from dual echo multislice gradient echo images covering the volume of interest.
Preclinical study: Scanner RF power and B 0 field optimisation for 13 C MR spectroscopy 13 C and 1 H power requirements for the RF coils were calibrated a similar approach to the human studies, employing a fiducial marker within the coil housing as a signal source for 13 C power calibration, and a volume-selective spectroscopy scan localised to the rat liver for 1 H power calibration.
Homogeneity of the scanner's B 0 magnetic field was performed by iterative manual adjustment of shim currents with concurrent measurement of water proton resonance linewidth from a 2 x 2 x 1 cm 3 volume of interest positioned over the liver. 1 H linewidth was 97 ± 19 Hz (mean ± standard deviation) for the preclinical studies.

Relative contributions of hepatic and non-hepatic tissue to 13 C spectra
A surface coil was used to localise the acquired spectra to liver tissue in our study.
Prior to human studies we performed measurements to determine the relative contribution of hepatic and non-hepatic tissue to the acquired signal. The active volume of the coil is principally liver, but also contains skin, subcutaneous fat, bone, and intercostal muscle. As the magnitude of the 13 C coil's B 1 field decreases with distance from the coil, a pulse power setting optimised for maximal hepatic signal results in an achieved tip angle greater than 90 degrees between the coil and the liver.
This results in T 1 -saturation immediately adjacent to the coil, with consequent reduction of the contribution of non-hepatic regions to the acquired spectrum. Based on our coil sensitivity measurements, on the morphology of study participants, and on the acquisition parameters used in the scan protocol, we estimate that hepatic tissue comprises 86 ± 4% by volume to the study spectra (ie. if a metabolite were distributed uniformly over hepatic and non-hepatic regions, 86% of the signal would arise from the liver (data not shown)). As non-hepatic tissue has lower glutathione content than liver, this further increases the dominance of the hepatic contribution. For these reasons we consider the contributions to the spectra from non-hepatic tissue to be sufficiently minor that they can be neglected in this study.
Acquisition of preclinical 13 C spectra to present a similar situation, aided further by the more ventral morphology of the rat liver (allowing coil positioning with minimal contribution from the ribcage and minimal separation between coil and liver). Little subcutaneous fat was observed adjacent to the coil, even in rats receiving the HFHC diet, and imaging prior to acquisition of 13 C spectra confirmed consistent positioning of the liver over the 13 C RF coil (Figure 1 F&H). Thus for preclinical studies the contributions of non-hepatic tissue to the spectra were assumed to be negligible.

Preclinical studies: data analysis and quantitation
Analysis of rat MR data was performed using jMRUI version 4.0 (Universitat Autònoma de Barcelona, Spain) (5,6). Spectral datasets were zero-filled from 512 to 1024 datapoints, 15 Hz of exponential line broadening was applied, then data were Fourier transformed and manually phased. The AMARES fitting algorithm (7)

Human study: Analysis of plasma glycine content and 13 C fractional enrichment
Perchloric acid extracts were prepared from plasma samples and extracts analysed

Preclinical study methods: Preparation and analysis of tissue extracts
Frozen liver samples were homogenized at room temperature in preweighed solutions of 50 mM phosphate buffer containing 25 mM monobromobimane. Proteins were precipitated by addition of perchloric acid and removed by centrifugation, then excess perchloric acid was removed by neutralisation with potassium hydroxide. The resultant thiol-bimane conjugate extracts were freeze dried for storage, then analysed by mass spectrometry using a Thermo Surveyor liquid chromatograph coupled to a Thermo LTQ linear ion trap mass spectrometer. The fractional enrichment of glutathione-bimane with 13 C label was determined as previously described (8).
Briefly, the isotope enrichment was calculated from the ratio of distributions of mass to charge ratio of 500, 499 and 498 amu by comparison to theoretical isotope distribution patterns calculated for glutathione-bimane using a mass spectrometry webtool (9). Glutathione concentration in the extracts was determined using HPLC analysis of glutathione-bimane conjugates. A portion of tissue extract samples was freeze dried, resuspended in D 2 O, and 1 H MR spectra acquired using a Bruker Avance III 500 MHz spectrometer equipped with a BBO probe for n=3 samples per experimental group.
Hepatic F 2 -isoprostane were isolated from liver tissue using the methods of Davies (10) and Morrow (11), comprising chloroform/methanol extraction, hydrolysis of esterified isoprostanes in phospholipids, and purification by cation exchange solid phase extraction. F 2 -isoprostane analysis was performed using a Prominence HPLC (Shimadzu, Kyoto, Japan) equipped with a C 18 column, coupled to a Q-Trap mass spectrometer (Sciex, Warrington, UK) operated using multiple reaction monitoring in negative ion mode, monitoring the transition of the precursor F 2 isoprostane ion (with a m/z of 353) to specific productions with m/z of 193, 127 and 115 for the F 2 -III, IV and VI isomers respectively.

Preclinical study: Histology
Formalin fixed liver sections were stained with Haematoxyin and Eosin (H&E) or 0.1% Sirius Red Picric solution following standard protocols. α-smooth muscle actin antibody immunohistochemistry (α-SMA) was performed on formalin-fixed liver sections as previously described (12). Photomicrographs were taken at x10 magnification on a Leica DMR microscope with a DFC 310 FX camera (Leica Microsystems, Wetzlar, Germany).