In vivo cortical glutathione response to oral fumarate therapy in relapsing-remitting multiple sclerosis: A single-arm open-label phase IV trial using 7-Tesla ¹H MRS

Highlights

• •This is an open-label, single-arm, single-center study on oral fumarates in multiple sclerosis.
• •¹H MRS showed increased prefrontal but not occipital glutathione with 12 months of fumarates.
• •Controls showed no such effects in brain glutathione concentration over 6 months.

Abstract

Background

This is an open-label, single-arm, single-center pilot study using 7-Tesla in vivo proton magnetic resonance spectroscopy (¹H MRS) to measure brain cortical glutathione concentration at baseline before and during the use of oral fumarates as a disease-modifying therapy for multiple sclerosis. The primary endpoint of this research was the change in prefrontal cortex glutathione concentration relative to a therapy-naïve baseline after one year of oral fumarate therapy.

Methods

Brain glutathione concentrations were examined by ¹H MRS in single prefrontal and occipital cortex cubic voxels (2.5 × 2.5 × 2.5 cm³) before and during initiation of oral fumarate therapy (120 mg b.i.d. for 7 days and 240 mg b.i.d. thereafter). Additional measurements of related metabolites glutamate, glutamine, myoinositol, total N-acetyl aspartate, and total choline were also acquired in voxels centered on the same regions. Seven relapsing-remitting multiple sclerosis patients (4 f / 3 m, age range 28–50 years, mean age 40 years) naïve to fumarate therapy were scanned at pre-therapy baseline and after 1, 3, 6 and 12 months of therapy. A group of 8 healthy volunteers (4 f / 4 m, age range 33–48 years, mean age 41 years) was also scanned at baseline and Month 6 to characterize ¹H-MRS measurement reproducibility over a comparable time frame.

Results

In the multiple sclerosis cohort, general linear models demonstrated a significant positive linear relationship between prefrontal glutathione and time either linearly across all time points (+0.05 ± 0.02 mM/month, t(27) = 2.6, p = 0.02) or specifically for factor variable Month 12 (+0.6 ± 0.3 mM/12 months, t(24) = 2.2, p = 0.04) relative to baseline. No such effects of time on glutathione concentration were demonstrated in the occipital cortex or in the healthy volunteer group. Changes in occipital total choline were further observed in the multiple sclerosis cohort as well as prefrontal total choline and occipital glutamine and myoinositol in the control cohort throughout the study duration.

Conclusions

While the open-label single-arm pilot study design and abbreviated control series cannot support firm conclusions about the influence of oral fumarate therapy independent of test–retest factors or normal biological variation in a state of either health or disease, these results do justify further investigation at a larger scale into the potential relationship between prefrontal cortex glutathione increases and oral fumarate therapy in relapsing-remitting multiple sclerosis.

1. Introduction

Multiple sclerosis (MS) is a chronic disorder of the central nervous system that leads to demyelination and neurodegeneration. Worsening of clinical symptoms despite stable magnetic resonance imaging (MRI) lesion burden implicates the involvement of processes other than acute inflammation and its immediate consequences (Lublin et al., 2014). To date, the cause of MS and the underlying neurobiological mechanisms leading to cell loss, brain atrophy, and progressive impairment remain unclear, and only partially effective treatments have been developed to prevent worsening disability and functional decline (Goldenberg, 2012).

Cells must rapidly increase their antioxidant capacity when challenged with oxidative stressors to counteract reactive oxygen species (ROS) and maintain homeostasis. Glutathione (GSH) is an endogenously synthesized antioxidant that plays a critical role in protecting cells against such oxidative damage (Coyle and Puttfarcken, 1993). In particular, glutathione provides a first line of defense against singlet oxygen and hydroxyl radicals that are known to cause cellular damage and eventual cell death (Bains and Shaw, 1997). Multiple lines of evidence suggest that oxidative stress is involved in MS pathology (Aoyama and Nakaki, 2013, Bains and Shaw, 1997). For example, ROS are generated by activated macrophages during inflammation and have been linked to damage of myelin, oligodendrocytes, and mitochondria (Paling et al., 2011). In addition, evidence of disrupted glutamate metabolism, itself a possible indication of oxidative stress, has been observed in contrast-enhancing MRI lesions (Srinivasan et al., 2010) and multiple sclerosis lesions of autopsied tissue (Werner et al., 2001). Glutathione is assumed to play a neuroprotective role through direct interaction with ROS or as part of enzymatic redox cycling by which ROS are rendered innocuous (Lu, 2013). Because glutathione is consumed in this protective process, its decrease is noted in brain tissue undergoing oxidative stress (Sohal, 2002). Upregulation of the Nuclear factor-erythroid factor 2-related factor 2 (Nrf2) signaling pathway under oxidative stress conditions is assumed to stimulate glutathione synthesis and restoration of redox homeostasis. Imbalanced glutathione metabolism has been hypothesized to play a role in MS tissue injury (Gonsette, 2008), and several studies have observed in vivo reductions in cortical glutathione in particularly progressive multiple sclerosis (Choi et al., 2018, Choi et al., 2017, Choi et al., 2011, Swanberg et al., 2021) that may be more pronounced in grey than white matter (Srinivasan et al., 2010).

It has been shown in primary cultures of cells of the central nervous system (CNS) that treatment with dimethyl fumarate (DMF), and its primary metabolite monomethyl fumarate (MMF), increases glutathione in a concentration-dependent manner (Scannevin et al., 2012). It has been hypothesized that DMF/MMF may exert these effects by facilitating nuclear translocation of Nrf2 to upregulate genes involved in protective actions against oxidative stress, including glutathione synthesis (Bomprezzi, 2015). To date, however, in vivo support for the involvement of glutathione in the therapeutic effect of fumarate therapy for multiple sclerosis has been lacking.

Tecfidera® (or BG-12) is an oral therapeutic agent containing DMF and MMF. Two randomized, placebo-controlled Phase 3 studies involving relapsing-remitting multiple sclerosis (RRMS) patients, CONFIRM (Fox et al., 2012) and DEFINE (Gold et al., 2012), demonstrated relapse reduction with Tecfidera® therapy by approximately 50% compared with placebo treatment. The drug was subsequently approved for treatment of RRMS in adults by the U.S. Food and Drug Administration in March 2013 (FDA, 2013).

Proton MR spectroscopy (¹H MRS) is the only quantitative method available to study glutathione metabolism in vivo in a non-invasive fashion. MRS enables the safe and repeated quantification of neurochemicals from the living brain. For MS, it allows the biochemical assessment of pathological changes and brain damage non-invasively from the earliest stage of the disease (De Stefano and Filippi, 2007). Several important challenges complicate the reliable quantification of glutathione and related metabolites like glutamate by ¹H MRS. The relatively low in vivo concentrations of glutathione (de Graaf, 2008), for instance, limit detection sensitivity; in addition, glutathione resonances overlap with and are therefore obscured by much stronger signals from creatine and macromolecules. Specialized ¹H-MRS methods like J-difference editing (JDE), a subtraction technique that extracts a metabolite of interest from the rest of the spectrum by exploiting a selected intramolecular coupling, must therefore be applied to reliably measure glutathione (Rothman et al., 1993).

Also methodologically difficult is the individual quantification of glutamate, an excitatory neurotransmitter and metabolic precursor to glutathione, as separate from its structurally similar metabolite glutamine. Increasing magnetic field strength improves the overall MRS detection sensitivity and enhances spectral dispersion, enabling the reliable separation of these two signals at 7 Tesla (Tkáč et al., 2001).

Using a one-hour 7-Tesla single-session ¹H-MRS scan protocol of validated reproducibility (Prinsen et al., 2017) we recently demonstrated that both prefrontal cortex glutathione and glutamate, as well as inhibitory neurotransmitter GABA, demonstrate MR-visible phenotype-specific changes in multiple sclerosis (Swanberg et al., 2021), in line with previous work predominantly conducted at lower field strengths less conducive to reliable isolation of glutamate (Swanberg et al., 2019). Voxel selection was based not on specific hypothesis of regional effect but rather the relative lack of previous ultra-high field research directed at the traditionally difficult-to-shim prefrontal cortex; the auxiliary occipital cortex voxel provided comparatively high expected data quality in addition to a means to assess the globality of any observed metabolic changes. Here, we apply similar methods over a longitudinal pilot study design to report the metabolic response of brain cortical glutathione concentration, in concert with those of glutamate and other related small molecules glutamine, myoinositol, total N-acetyl aspartate, and total choline, to Tecfidera® therapy in patients with relapsing-remitting multiple sclerosis.

2. Methods

2.1. Study design and endpoints

This was an open-label, single-arm, single-center pilot study looking at cortical glutathione at baseline before initiation of oral fumarate therapy (Month 0; M₀) and compared to cortical glutathione at Month 1 (M₁), Month 3 (M₃), Month 6 (M₆) and Month 12 (M₁₂). Scan M₀ was performed once all Tecfidera® paperwork, authorization, and pharmacy request were completed. Early scans (M₁-M₆) were performed within a week of each attempted time point, while the last time point M₁₂ was performed between months 10 and 12. Seven patients with relapsing-remitting multiple sclerosis were included (4 f / 3 m; 28–50 years old, mean age 40 years). Therapy compliance was assessed at all study time points. A control group of 8 healthy volunteers was also included (4 f / 4 m; age range 33–48 years, mean age 41 years) (Table 1). These were studied twice, at M₀ and M₆, enabling a test–retest assessment of the established experimental procedure (Fig. 1, Fig. 2).

Table 1.

Study participant demographics.

Group	Sex	Age at baseline (years)	Age at diagnosis (years)	EDSS
MS	female	45	30	1.5
MS	female	28	20	1.5
MS	female	50	49	1.5
MS	male	42	36	2.5
MS	male	46	45	–
MS	female	34	18	1
MS	male	32	32	–
HC	female	45
HC	male	48
HC	female	44
HC	female	33
HC	female	39
HC	male	46
HC	male	31
HC	male	49

MS: multiple sclerosis; HC: healthy control; EDSS: Expanded Disability Status Scale.

Fig. 1.

Study design for investigating the metabolic response to Tecfidera® therapy in patients with relapsing-remitting multiple sclerosis. Relapsing-remitting multiple sclerosis (RR-MS) patients were scanned four times (M₁, M₃, M₆, and M₁₂) following a pre-treatment baseline scan (M₀). In parallel, age- and sex-matched healthy controls were scanned twice (M₀, M₆). J-difference-edited glutathione (GSH) spectra are presented in the schematic for RR-MS patients, while STimulated Echo Acquisition Mode (STEAM) spectra are presented for the healthy control schematic; in reality both experimental groups received both types of scan at each time point. RR-MS: Relapsing-remitting multiple sclerosis; NAA: N-acetyl aspartate; ppm: parts per million.

Fig. 2.

Anatomical positioning for 7-Tesla single-voxel proton spectroscopy acquisition and CONSORT diagram for study participant attrition. A. Example of a T₁-weighted anatomical image in axial, coronal, and sagittal planes used for voxel positioning (voxels 2.5x2.5x2.5 cm³) on the prefrontal cortex (PFC) and occipital cortex (OCC) in a relapsing-remitting multiple sclerosis patient at Month 0. B. No unexpected adverse events were reported for the duration of the study. Following two dropouts (N = 1 patient and N = 1 control) after time point 0, one relapsing-remitting multiple sclerosis patient missed the last time point (M₁₂) and was studied only during months zero through six (M₀-M₆). Excepting this point, full data sets representing the concentrations of glutathione and associated biochemicals were acquired in the prefrontal cortex for all 7 patients and all 8 control individuals. Due to time constraints, a similar metabolic profile was acquired in the occipital cortex from 6 patients and 6 control individuals, minus time point M₆ from one patient and M₁₂ from two. R: right; P: posterior; I: inferior; GSH: glutathione; STEAM: STimulated Echo Acquisition Mode; OCC: occipital cortex.

The primary objective of this pilot study was to directly estimate prefrontal cortex brain glutathione concentrations in vivo using ¹H MRS at 7 T before and after initiation of Tecfidera® in established MS patients treatment-naïve (first line) or switching therapy (Table 2). Secondary objectives included assessing changes in occipital glutathione as well as changes in related metabolites glutamate, glutamine, myoinositol, N-acetyl aspartate, and total choline in both prefrontal and occipital cortices over the same period.

Table 2.

Multiple sclerosis participant disease-modifying therapy (DMT) history.

	DMT start	DMT stop	DMT start	DMT stop	DMT start	DMTstop	DMT start	DMT stop	DMT start
1	interferon β-1a (Avonex®) Sept. 2007	Sept. 2014	dimethyl, monomethyl fumarate (Tecfidera®) Oct. 2014
2	glatiramer acetate (Copaxone®) Aug. 2007	Aug. 2008	interferon β-1a (Rebif®) Sept. 2008	Mar. 2009	natalizumab (Tysabri®) Jul. 2010	Nov. 2012	natalizumab (Tysabri®) Oct. 2013	Jul. 2014	dimethyl, monomethyl fumarate (Tecfidera®) Oct. 2014
3	dimethyl, monomethyl fumarate (Tecfidera®) Nov. 2014
4	glatiramer acetate (Copaxone®) Jan. 2014	Jan. 2015	dimethyl, monomethyl fumarate (Tecfidera®) Jan. 2015
5	dimethyl, monomethyl fumarate (Tecfidera®) Jan. 2015
6	interferon β-1a (Avonex®) Sept. 2009	Jan. 2015	dimethyl, monomethyl fumarate (Tecfidera®) Feb. 2015
7	dimethyl, monomethyl fumarate (Tecfidera®) Mar. 2015

2.2. Participant recruitment

Key eligibility criteria for multiple sclerosis patients included 1) 18–55 years of age inclusive; 2) diagnosis of relapsing-remitting multiple sclerosis evaluated at the Yale MS Center in accordance with the revised McDonald criteria (Lublin and Reingold, 1996, Polman et al., 2011); 3) naïve to MS therapy or switching from an FDA-approved MS therapy including IFN-β formulations, Copolymer-1, teriflunomide, and fingolimod to Tecfidera®; 4) Expanded Disability Status Scale (EDSS) score 0 to 5.5 inclusive. Candidates were excluded from study entry if they 1) presented with primary progressive multiple sclerosis (PPMS); 2) had previous exposure or known allergies to fumarates; 3) were switching from cyclophosphamide or mitoxantrone to Tecfidera®; 4) had contraindications for magnetic resonance scanning; 5) exhibited a known presence of other neurological disorders which may mimic multiple sclerosis; 6) were pregnant or lactating; 7) required chronic treatment with systemic corticosteroids or immunosuppressants during the course of the study; 8) had a history of or currently active primary or secondary immunodeficiency; 9) presented with active infection, or history of or known presence of recurrent or chronic infection (e.g. hepatitis B or C, HIV, syphilis, tuberculosis); 10) had a history of progressive multifocal leukoencephalopathy; or 11) had apparent contraindications to or intolerance of oral or intravenously (iv) applied corticosteroids. In addition, no other disease-modifying treatment was permitted during the course of the study. The pharmacological management of symptoms, including dietary supplementation, was allowed, but subjects were advised not to take dietary glutathione supplements to prevent potential confounding effects. Patients were allowed to receive steroid therapy during acute clinical exacerbations, and subsequent scans would take place 4 weeks after the last steroid administration. However, no clinical exacerbations were observed before MRS investigations, and subsequent rescheduling was not necessary. Disability outcome measures were not collected throughout this pilot study.

2.3. Spectral data acquisition

All MR imaging (MRI) and spectroscopy (MRS) experiments were performed at the ultra-high-field 7-Tesla MR facility at the MR Research Center (MRRC) of Yale University. All participants provided written informed consent, and study procedures were conducted in accordance with Yale University Institutional Review Board (IRB) guidelines for human-subjects research. Scanning was carried out on a 7-Tesla head-only MR system (Agilent, Santa Clara, CA, USA) interfaced to a DirectDrive spectrometer operating at 298.1 MHz for protons. The MR scanner was equipped with custom-designed actively shielded gradients (Magnex Scientific, Oxford, UK) and operated with Vnmrj 2.3A software (Varian, Santa Clara, CA, USA). An 8-channel transmit-receive array was used for spin handling and signal reception. Reproducibility of the employed methods has been validated (Prinsen et al., 2017) and their potential as a clinical research tool reported in a case-control analysis of cortical small-molecule metabolic abnormality in multiple sclerosis (Swanberg et al., 2021).

T₁-weighted anatomical images were obtained with a customized inversion-recovery prepared MRI sequence (field of view 200x220x78 mm³, matrix size 256x256x39, resolution 0.78x0.86x2.00 mm³, inversion time T_I 1000 ms, repetition time T_R 3000 ms, echo time T_E 6 ms, acquisition time 5 min). The anatomical images were used in the first session to position the spectroscopy voxels (2.5x2.5x2.5 cm³) on the midline prefrontal or occipital cortex (Fig. 2A). The voxel was visually repositioned in subsequent sessions.

B₀ shimming included zero- through third-order spherical harmonic terms and was calculated with customized software (B0DETOX (Juchem et al., 2015)). B₁ phase shimming was achieved through in-house developed MR methods and software (IMAGO (Prinsen et al., 2014)). The B₀ magnetic field of the unshimmed brain was derived from five single-echo gradient-echo images (field of view 220x220x60 mm³, matrix 126x64x20, T_E 3.8 ms, T_R 1.3 s) at additional T_E delays of 0/0.2/0.5/1.5/3 ms.

Glutathione measurement was achieved with a customized semi-Localization by MEscher-GARwood Adiabatic SEelective Refocusing (MEGA-sLASER) J-difference editing (JDE) sequence as described previously (Prinsen et al., 2017, Swanberg et al., 2018, Swanberg et al., 2021) (T_R 3 s, T_E 72 ms, 64 averages per JDE condition, acquisition time 7 min). The editing pulse was applied at 4.56 ppm to select the J-coupled CH₂ signal from the 2.95-ppm glutathione cysteine moiety and far away (5 kHz offset) for the non-edited condition. Water suppression was achieved with a CHemical Shift Selective (CHESS) module. Water references were acquired for eddy-current correction (Klose, 1990) (Fig. 3A).

Fig. 3.

Spectral data acquisition for glutathione, glutamate, and related metabolites at 7 Tesla. J-difference editing (JDE) of glutathione (GSH) consisting of an edited condition (A, blue) and a non-inverted reference (A, violet) was performed using MEscher-GArwood semi-Localization by Adiabatic SElective Refocusing (MEGA-sLASER; voxel size 2.5x2.5x2.5 cm³, T_R 3 s, T_E 72 ms, 64 averages per JDE condition, acquisition time 7 min). The difference spectrum (black) exhibits expected co-edited NAA at 2.49 and 2.67 ppm and allows the isolation of the GSH CH₂ signal of the cysteine moiety at 2.95 ppm (dotted vertical line). STimulated Echo Acquisition Mode (STEAM; voxel size 2.5x2.5x2.5 cm³, T_R 3 s, T_E 10 ms, mixing time 50 ms, 96 averages, acquisition time 5 min) was employed in order to measure glutamate and related proton magnetic resonance spectroscopy-visible molecules in an overlapping, albeit smaller, voxel. Spectra from a single relapsing-remitting multiple sclerosis patient at Month 12. Quantified metabolites shown in violet; others fit but not reported depicted in grey. tCr: total creatine; NAA: N-acetyl aspartate; Glu: glutamate; Gln: glutamine; tCho: total choline; PE: phosphorylethanolamine; sIns: scylloinositol; mIns: myoinositol; Glx: glutamate + glutamine; GABA: γ-aminobutyric acid; Tau: taurine; Glc: glucose; Gly: glycine; Asc: ascorbate; Asp: aspartate; ppm: parts per million.

Lastly, glutamate, glutamine, myoinositol, total N-acetyl aspartate (N-acetyl aspartate + N-acetyl aspartylglutamate), and total choline (choline + phosphocholine + glycerophosphocholine) were measured using the short echo-time STimulated Echo Acquisition Mode (STEAM) sequence (Frahm et al., 1989) (T_R 3 s, T_E 10 ms, mixing time 50 ms, 96 averages, acquisition time 5 min). Macromolecule resonances were suppressed by a non-selective inversion pulse (T_I 320 ms). Water suppression was based on CHESS, outer-volume suppression was used for improved localization specificity (Juchem et al., 2007), and water references were again acquired for eddy-current correction (Fig. 3B).

2.4. Spectral data processing and quantification

Spectral processing and quantification were adapted from procedures previously detailed (Swanberg et al., 2021). Briefly, spectra from individual acquisitions and receivers were acquired and stored separately. Spectral processing was achieved in MATLAB-based software INSPECTOR (Gajdosik et al., 2021, Prinsen et al., 2014). First, eddy-current phase correction (Klose, 1990) and sensitivity-weighted summation of the receive channels were applied. Spectra were phase- and frequency-aligned individually using a least-squares fit and a cross-correlation algorithm, respectively. Finally, the phase- and frequency-aligned spectra from individual acquisitions were averaged.

Metabolite quantification was achieved in LCModel (Provencher, 1993) using density-matrix-simulated basis functions generated in SpinWizard (de Graaf et al., 2015) based on spectral characteristics summarized previously (Govindaraju et al., 2000). The STEAM basis set included simulated spectra of ascorbate, aspartate, choline, creatine, GABA, glycerophosphocholine, glutathione, glucose, glutamate, glutamine, glycine, myoinositol, N-acetyl aspartate, N-acetyl aspartylglutamate, phosphocholine, phosphocreatine, phosphorylethanolamine, scylloinositol, and taurine. The MEGA-sLASER difference spectrum basis set included basis functions for only glutathione and NAA because all other peaks were subtracted out. MEGA-sLASER edit-off condition basis functions included aspartate, choline, creatine, GABA, glutathione, glutamate, glutamine, myoinositol, N-acetyl aspartate, N-acetyl aspartylglutamate, phosphocholine, phosphocreatine, phosphorylethanolamine, and scylloinositol. All metabolites were quantified according to a constant 10 mM total creatine (creatine + phosphocreatine) referenced from the same acquisition sequence as that metabolite (Fig. 4).

Fig. 4.

Spectral quantification for glutathione, glutamate, and related metabolites at 7 Tesla. Quantification of metabolite signatures present in the MEscher-GARwood semi-Localization by Adiabatic SElective Refocusing (MEGA-sLASER) difference spectrum for glutathione (GSH; upper left), the corresponding J-difference-editing (JDE)-edit-off MEGA-sLASER spectrum used to calculate the creatine reference (lower left), and STimulated Echo Acquisition Mode (STEAM) spectrum from the same voxel used to measure glutamate and related metabolites (right) proceeded by linear combination modeling in LCModel of basis spectra density-matrix-simulated in SpinWizard to reflect the specific conditions of the pulse sequences used. Here basis spectra are shown in blue, the regularized cubic spline baseline underlying the fit in green, and fit residual (acquired data, in black, minus the fit model, in red) in grey. Seven metabolites, shown in violet, were considered in subsequent analysis: glutathione, glutamate, glutamine, myoinositol, total N-acetyl aspartate (tNAA, the sum of N-acetyl aspartate and N-acetyl aspartylglutamate), total choline (tCho, the sum of choline, glycerophosphocholine, and phosphocholine), and total creatine (tCr, the sum of creatine and phosphocreatine). Total creatine was set to 10 mM and used as a concentration reference for metabolites from the corresponding acquisitions such that MEGA-sLASER-measured glutathione was referenced to tCr from corresponding MEGA-sLASER edit-off spectra while all other STEAM-acquired metabolites were referenced to tCr from STEAM. GABA: γ-aminobutyric acid; ppm: parts per million.

Spectral quality was assessed as full width at half maximum (FWHM) of the 2.01-ppm N-acetyl aspartate singlet as well as the 3.03-ppm total creatine singlet in the non-inverted JDE condition for MEGA-sLASER and as 3.03-ppm total creatine for STEAM. LCModel quantification precision for each metabolite signal of reported concentration was estimated via Cramér-Rao lower bounds (CRLB) (Cavassila et al., 2000).

2.5. Statistical analysis

All group statistics were calculated in R (v. 4.0.5, R Foundation for Statistical Computing, Vienna, Austria) (Team.). Descriptive statistics are reported as mean ± standard deviations and linear model coefficients as mean ± standard errors unless otherwise noted. Longitudinal changes in metabolite concentration were assessed individually within multiple sclerosis and healthy control groups using generalized linear mixed models (Bates et al., 2015) with random variable intercept for each individual under study and either numeric or factor coefficients for the variable time (M₀ and M₆ for control and M₀, M₁, M₃, M₆, and M₁₂ for multiple sclerosis); t-test statistics for model coefficients employed the Satterthwaite approximation for degrees of freedom estimation (Kuznetsova et al., 2017). Initial restricted-maximum-likelihood models employed a linear linking function with assumption of normal error; log-linked maximum-likelihood models with assumed Gamma error distributions (Venables et al., 2002) were instead reported in the case of a significant Shapiro-Wilk test statistic for non-normal residuals on the former. Base α was set to 0.05.

3. Results

3.1. Participant attrition

No unexpected adverse events were reported for the duration of the study. Real-time data screening during experiment execution was applied in all sessions to assess attained spectral quality and for a preliminary MRS data analysis, and MRS investigation of one patient and one healthy control was repeated because of excessive movement during data acquisition. The patient reported flushing as the underlying reason for excessive movement during the study (1-month time point). Notably, this individual had taken the drug shortly before the MRS session, and flushing is a reported side effect of Tecfidera® therapy. Leaving more time between the administration of the drug and the MRS session prevented the problem in the remaining sessions (Months 3–12).

One MS patient missed the last time point (M₁₂) and was studied only during Months 0 through 6 (M₀-M₆). Otherwise, full data sets representing the concentrations of glutathione and associated biochemicals were acquired in the prefrontal cortex for all 7 MS patients and all 8 healthy control individuals. Due to time constraints and primary endpoint definition in the prefrontal cortex, a similar metabolic profile was acquired in the occipital cortex from 6 patients and 6 healthy control individuals, minus time point M₆ from one patient and M₁₂ from two.

3.2. Data quality

The main target metabolite glutathione as well as secondary target glutamate were detectable in all completed acquisitions. Average 3.03-ppm total creatine singlet line widths for STEAM of 12.7–16.5 Hz (median 13.8 Hz) and for non-inverted MEGA-sLASER of 11.6–15.2 Hz (median 12.8 Hz) were found at each time point, group, and voxel (Table 3). Besides a significant effect of linear model fixed factor time M₃ on both occipital creatine (−0.2 ± 0.06 Hz/3 months, t(15) = -2.6, p = 0.02) and N-acetyl aspartate (−2.3 ± 0.9 Hz/3 months, t(16) = -2.5, p = 0.02) singlet FWHM in the multiple sclerosis group, MEGA-sLASER spectral line widths exhibited stability over time in both groups and voxels (Table3). Spectral reproducibility for both voxels and acquisition types showed no visibly apparent artefactual differences across time points (Fig. 5; Supplementary Figure 1).

Table 3.

Spectral full width at half maximum (FWHM; Hz) by voxel, group, and time point.

	PREFRONTAL		OCCIPITAL
	MS (N = 6–7)	HC (N = 8)	MS (N = 4–6)	HC (N = 6)
STEAM tCr	a,b		a,b	a,b
0	16.5 ± 6.5	13.1 ± 1.9	14.5 ± 2.2	12.7 ± 1.2
1	14.6 ± 3.8		12.8 ± 0.8
3	14.6 ± 2.9		13.0 ± 1.9
6	13.1 ± 2.8†	13.8 ± 2.7	13.7 ± 1.8	14.3 ± 3.3
12	14.3 ± 2.9		13.4 ± 1.8
sLASER tCr	a		a,b
0	13.2 ± 2.6	12.3 ± 2.9	13.4 ± 2.1	12.0 ± 0.9
1	15.2 ± 5.3		12.2 ± 0.5
3	12.9 ± 2.7		11.5 ± 0.8*
6	11.6 ± 2.4	12.7 ± 1.5	12.5 ± 1.7	12.8 ± 2.1
12	13.4 ± 0.9		12.0 ± 1.0†
sLASER tNAA	a		a
0	12.2 ± 2.7	11.7 ± 2.1	13.4 ± 2.1	11.8 ± 1.6
1	15.0 ± 6.7		12.2 ± 0.8
3	12.0 ± 2.2		11.1 ± 1.1*
6	10.7 ± 2.2	12.4 ± 2.0	12.6 ± 2.0	12.0 ± 2.0
12	12.5 ± 1.1		11.5 ± 1.3†

t-test for general(ized) linear mixed model factor or numeric (equivalent for control models) time coefficient **p < 0.01, *p < 0.05, †p < 0.1. Numeric-time model results shown in series header; factor-time model results displayed inline.

^a Numeric time variable general linear mixed model residuals non-normal; Gamma mixed model with log link reported.

^b Factor time variable general linear mixed model residuals non-normal; Gamma mixed model with log link reported.

Fig. 5.

Prefrontal cortex spectra. Excepting a missing data set from Month 12 (M₁₂) in one relapsing-remitting multiple sclerosis (MS) patient participant, full metabolic profiles including glutathione, glutamate, glutamine, total N-acetyl aspartate, myoinositol, and total choline were obtained from the prefrontal cortex (PFC) of 8 control (HC) and 7 multiple sclerosis patient participants. Shown are the mean (black) spectra over the full time series (M₀ and M₆ for control; M₀, M₁, M₃, M₆, and M₁₂ for multiple sclerosis) for each participant. Violet (J-difference-edited spectra for glutathione) and blue (STEAM spectra for the other metabolites) bands demonstrate the mean plus and minus one standard deviation at each frequency point over the time series for each individual. Spectra were quantified with minimal preprocessing for baseline control, as is apparent in the average spectrum for, e.g. the first multiple sclerosis patient, dominated by suboptimal water suppression for N_m = 0. N_m = Month number.

LCModel fits yielded average Cramér-Rao Lower Bound values of 10.2–13.7% (median 10.9%) for glutathione, ≤2% for glutamate, ≤4% for glutamine, and <5% for most additional metabolites (<25% in all cases). Prefrontal glutathione Cramér-Rao Lower Bounds demonstrated a significant effect of numeric (-0.03 ± 0.009%/month, t(26) = -3.2, p = 0.003) or factor (−0.3 ± 0.09%/6 months, t(23) = −3.1, p = 0.005; −0.3 ± 0.0.1%/12 months, t(23) = -2.8, p = 0.01) variable time relative to baseline in the multiple sclerosis group (Table 4).

Table 4.

Metabolite Cramér-Rao Lower Bounds (CRLB; %) by voxel, group, and time point.

	PREFRONTAL		OCCIPITAL
	MS (N = 6–7)	HC (N = 8)	MS (N = 4–6)	HC (N = 6)
GSH	**|a,b		†
0	13.7 ± 3.4	12.9 ± 4.2	10.2 ± 0.8	10.5 ± 2.1
1	13.6 ± 4.3		10.5 ± 1.2
3	12.3 ± 1.7		10.8 ± 1.2
6	10.3 ± 1.1**	11.0 ± 1.5	10.8 ± 2.2	13.5 ± 4.3
12	10.7 ± 1.2*		12.3 ± 3.2†
tCr (sLASER)‡	a	a,b	a,b	a,b
0	2.7 ± 0.8	2.5 ± 0.5	2.0 ± 0	2.3 ± 0.5
1	2.6 ± 0.8		2.7 ± 0.8†
3	3.0 ± 0.6		2.2 ± 0.4
6	2.3 ± 0.5	2.4 ± 0.5	2.2 ± 0.4	2.2 ± 0.4
12	3.2 ± 0.8		1.8 ± 0.5
Glu
0	1.9 ± 0.4	2 ± 0	2 ± 0	1.8 ± 0.4
1	1.9 ± 0.4		2 ± 0
3	1.9 ± 0.4		2 ± 0
6	1.9 ± 0.4	2 ± 0	2 ± 0	1.8 ± 0.4
12	2.0 ± 0		2 ± 0
Gln	a,b		a
0	3.6 ± 0.8	3.8 ± 0.9	3.7 ± 0.5	3.8 ± 0.4
1	3.3 ± 0.5		3.8 ± 0.4
3	3.3 ± 0.5		3.6 ± 0.5
6	3.3 ± 0.5	3.6 ± 0.5	3.6 ± 0.5	3.8 ± 0.8
12	3.3 ± 0.5		3.5 ± 0.6
tNAA
0	1.3 ± 0.5	1.3 ± 0.5	1.2 ± 0.4	1 ± 0
1	1.0 ± 0		1.0 ± 0
3	1.0 ± 0		1.0 ± 0
6	1.1 ± 0.4	1.1 ± 0.4	1.0 ± 0	1 ± 0
12	1.2 ± 0.4		1.0 ± 0
tCho1	a,b
0	14.4 ± 4.6	21.1 ± 5.9	12.7 ± 6.6	19.3 ± 7.6
1	12.0 ± 2.6		13.7 ± 9.6
3	20.3 ± 12.6†		14.8 ± 4.7
6	14.7 ± 3.5	18.4 ± 4.3	13.0 ± 6.0	14.0 ± 6.6
12	12.5 ± 6.1		13.5 ± 5.0
mIns	a,b	a,b
0	3.4 ± 1.8	3.0 ± 0	3.0 ± 0.6	3.2 ± 0.8
1	2.9 ± 0.7		2.8 ± 0.8
3	2.9 ± 0.4		2.6 ± 0.5
6	2.9 ± 0.4	3.3 ± 0.5	2.6 ± 0.5	3.3 ± 0.8
12	3.0 ± 0.6		3.3 ± 0.5
tCr (STEAM)‡	a,b	a,b	a,b	a,b
0	1.3 ± 0.5	1.4 ± 0.5	1.2 ± 0.4	1.0 ± 0.0
1	1.1 ± 0.4		1.0 ± 0
3	1.0 ± 0†		1.0 ± 0
6	1.0 ± 0†	1.4 ± 0.5	1.0 ± 0	1.2 ± 0.4
12	1.2 ± 0.4		1.3 ± 0.5

t-test for general(ized) linear mixed model factor or numeric (equivalent for control models) time coefficient **p < 0.01, *p < 0.05, †p < 0.1. Numeric-time model results shown in series header; factor-time model results displayed inline. GSH: glutathione; tCr: total creatine; Glu: glutamate; Gln: glutamine; tNAA: total N-acetyl aspartate; tCho: total choline; mIns: myoinositol.

¹ Reported as CRLB for choline + glycerophosphocholine.

‡Unitless reference signal.

^a Numeric time variable general linear mixed model residuals non-normal; Gamma mixed model with log link reported.

^b Factor time variable general linear mixed model residuals non-normal; Gamma mixed model with log link reported.

A significant positive effect of time was demonstrated for prefrontal cortex creatine signal magnitude in both the sLASER and STEAM experiments in the multiple sclerosis group; a similarly negative effect was demonstrated in the occipital cortex. A negative effect of time on occipital STEAM creatine only was demonstrated in the control cohort (Table 5).

Table 5.

Metabolite concentration (mM) by voxel, group, and time point.

	PREFRONTAL		OCCIPITAL
	MS (N = 6–7)	HC (N = 8)	MS (N = 4–6)	HC (N = 6)
GSH (mM)	*		a,b
0	2.0 ± 0.6	1.9 ± 0.4	1.8 ± 0.4	1.9 ± 0.3
1	2.1 ± 0.4		2.0 ± 0.5
3	2.2 ± 0.4		1.9 ± 0.3
6	2.3 ± 0.4	2.1 ± 0.3	1.8 ± 0.4	1.5 ± 0.4
12	2.6 ± 0.7*		1.6 ± 0.2
tCr (sLASER; a.u.)‡	*		*	†
0	5401 ± 1109	5486 ± 829	5650 ± 794	5587 ± 894
1	4931 ± 919		5287 ± 643
3	5654 ± 1129		4782 ± 594**
6	5847 ± 734	5693 ± 607	5002 ± 669*	4762 ± 769†
12	5547 ± 1008		4610 ± 822**
Glu (mM)	a,b	a,b
0	12.1 ± 1.3	11.4 ± 0.8	10.7 ± 1.1	10.5 ± 0.7
1	11.8 ± 0.6		10.9 ± 1.0
3	11.6 ± 0.6		10.6 ± 0.8
6	11.7 ± 0.5	11.7 ± 0.9	10.3 ± 0.4	10.7 ± 0.8
12	11.5 ± 0.6		10.6 ± 0.8
Gln (mM)	a,b			*
0	4.9 ± 0.8	4.5 ± 0.7	4.1 ± 0.5	3.7 ± 0.3
1	4.8 ± 0.3		4.0 ± 0.2
3	4.6 ± 0.2		4.1 ± 0.4
6	4.6 ± 0.4	4.7 ± 0.6	3.9 ± 0.6	4.0 ± 0.3*
12	4.8 ± 0.5		4.3 ± 0.5
tNAA (mM)	a,b
0	12.2 ± 1.3	11.8 ± 0.7	13.4 ± 0.9	13.3 ± 1.1
1	12.0 ± 0.9		13.5 ± 1.3
3	11.8 ± 0.7		13.7 ± 0.9
6	12.0 ± 0.9	12.1 ± 0.9	13.4 ± 0.9	13.4 ± 0.5
12	11.7 ± 0.6		13.3 ± 1.1
tCho (mM)	b	*	a,b
0	1.9 ± 0.2	1.8 ± 0.1	1.4 ± 0.4	1.0 ± 0.2
1	2.0 ± 0.2		1.2 ± 0.1*
3	1.9 ± 0.1		1.1 ± 0.1**
6	1.9 ± 0.1	1.7 ± 0.1*	1.1 ± 0.1*	1.1 ± 0.2
12	1.9 ± 0.1		1.2 ± 0.1†
mIns (mM)				*
0	8.3 ± 1.4	8.1 ± 0.7	8.3 ± 0.9	7.1 ± 0.7
1	8.5 ± 1.3		8.1 ± 1.1
3	8.5 ± 1.0		7.7 ± 0.3
6	8.7 ± 1.3†	8.1 ± 0.8	8.0 ± 1.0†	7.4 ± 0.7*
12	8.8 ± 1.6		8.3 ± 0.9
tCr (STEAM; a.u.)‡	**		*	*
0	4.5 ± 1.2	4.5 ± 0.9	4.6 ± 0.8	4.6 ± 0.9
1	4.2 ± 0.9		4.3 ± 0.5
3	4.7 ± 1.0		4.3 ± 0.4
6	4.8 ± 0.9	4.8 ± 1.0	4.3 ± 0.4	3.9 ± 0.8*
12	5.0 ± 1.1*		3.8 ± 0.9*

t-test for general(ized) linear mixed model factor or numeric (equivalent for control models) time coefficient **p < 0.01, *p < 0.05, †p < 0.1. Numeric-time results shown in series header; factor-time model results displayed inline. GSH: glutathione; tCr: total creatine; Glu: glutamate; Gln: glutamine; tNAA: total N-acetyl aspartate; tCho: total choline; mIns: myoinositol.

‡Unitless reference signal.

^a Numeric time variable general linear mixed model residuals non-normal; Gamma mixed model with log link reported.

^b Factor time variable general linear mixed model residuals non-normal; Gamma mixed model with log link reported.

3.3. Alterations in brain glutathione during study duration

Prefrontal glutathione exhibited a significant positive effect of numeric (+0.05 ± 0.02 mM/month, t(27) = 2.6, p = 0.02) or factor (+0.6 ± 0.3 mM/12 months, t(24) = 2.2, p = 0.04) variable time relative to baseline in the multiple sclerosis group. No effects of time on glutathione concentration were found in the occipital cortex or in the healthy control group (Table 5; Fig. 6; Supplementary Fig. 2).

Fig. 6.

Prefrontal glutathione exhibited temporal increases in multiple sclerosis patients but not control participants. Multiple sclerosis (MS) participants demonstrated a significant effect of time on prefrontal glutathione concentration from baseline to Month 12 (M₁₂) (as numeric variable: +0.05 ± 0.02 mM/month, t(27) = 2.6, p = 0.02; as factor variable: +0.6 ± 0.3 mM/12 months, t(24) = 2.2, p = 0.04). No effect of time on prefrontal glutathione was seen in control (HC) participants. By contrast, controls demonstrated a negative effect of time on prefrontal total choline from baseline to Month 6 (M₆) (as numeric: −0.01 ± 0.004 mM/month, t(7) = -3.4, p = 0.01; equivalently as factor: −0.08 ± 0.02 mM/6 months, t(7) = -3.4, p = 0.01). GSH: glutathione; Glu: glutamate; Gln: glutamine; tNAA: total N-acetyl aspartate (N-acetyl aspartate + N-acetyl aspartylglutamate); tCho: total choline (choline + phosphocholine + glycerophosphocholine); mIns: myoinositol. T-test significance for numeric-time model coefficients shown in series header; those for factor-time model coefficients displayed on the plot field: *p < 0.05; **p < 0.01.

3.4. Alterations in overall brain biochemistry during study duration

Occipital total choline demonstrated a significant negative effect on factor variable time at M₁ (−0.2 ± 0.07 mM/1 month, t(16) = −2.4, p = 0.03), M₃ (−0.2 ± 0.07 mM/3 months, t(16) = −3.2, p = 0.006), and M₆ (−0.2 ± 0.07 mM/6 months, t(16) = −2.5, p = 0.03) relative to baseline in the multiple sclerosis group. No other significant effects of time on metabolite concentration were seen for this cohort in either voxel.

In the healthy control group, time exerted a significant effect on prefrontal total choline (as numeric: −0.01 ± 0.004 mM/month, t(7) = -3.4, p = 0.01; equivalently as factor: −0.08 ± 0.02 mM/6 months, t(7) = -3.4, p = 0.01), occipital glutamine (+0.05 ± 0.01 mM/month, t(5) = 4.0, p = 0.01; equivalently as factor: +0.3 ± 0.07 mM/6 months, t(5) = 4.0, p = 0.01), and occipital myoinositol (+0.04 ± 0.010 mM/month, t(5) = 4.0, p = 0.01; equivalently as factor: +0.2 ± 0.06 mM/6 months, t(5) = 4.0, p = 0.01) (Table 5; Fig. 6; Supplementary Fig. 2).

4. Discussion

In this open-label, single-arm, single-center pilot study on the effect of oral fumarate therapy on cortical glutathione in relapsing-remitting multiple sclerosis, glutathione and other neurochemicals key to MS pathology were quantified in the prefrontal and occipital cortex with ¹H MRS at 7 Tesla over the course of newly initiated Tecfidera® therapy. Single-voxel MRS was used as opposed to magnetic resonance spectroscopic imaging as performed in previous in vivo studies of glutathione in multiple sclerosis (Choi et al., 2018, Choi et al., 2017, Choi et al., 2011, Srinivasan et al., 2010) as B₀ and B₁ conditions can be highly optimized and reasonably short individual acquisition times can be realized.

High-quality spectra characterized by a flat baseline, narrow spectral lines, and minimal residual water and macromolecule resonances were obtained. These observations are substantiated by the full width at half maximum values for creatine and/or N-acetyl aspartate singlets reported, as well as low Cramér-Rao Lower Bound values for all metabolites under investigation. Except for a decrease in relative prefrontal glutathione Cramér-Rao Lower Bounds expected for observed increases in glutathione concentration for this group and voxel, these quality measures were largely consistent across time points for both groups under study and do not suggest systematic methodological confounds over the study duration in either voxel or cohort.

As per current international consensus recommendations for single-voxel spectroscopy (Öz et al., 2021), the sLASER sequence with spectral editing for glutathione was chosen for glutathione measurement due to superior voxel definition and minimization of anomalous phasing between weakly coupled resonances by chemical shift displacement errors, particularly important considerations at higher field strengths like 7 Tesla (Lange et al., 2006, Öz et al., 2021). Under this acquisition scheme, the primary finding of this work was an increase in prefrontal glutathione concentration in the multiple sclerosis group over the time course of the study, as modeled either by time as a continuous numeric variable or as a discretized factor variable for each time point after baseline (Months 1, 3, 6, and 12 independently). No change was seen in glutathione for the occipital cortex in either group or either voxel in the healthy control group. Reported glutathione concentrations determined in the human occipital cortex at 7 Tesla have ranged from 0.5 mM to 2.3 mM (e.g., (Emir et al., 2012, Lin et al., 2012, Marjanska et al., 2012, Terpstra et al., 2010)). The cortical glutathione concentrations presented in the current study fall into the upper end of this range (1.5–2.1 mM) for healthy controls. Furthermore, they are comparable for the relapsing-remitting multiple sclerosis patients as per previous cross-sectional null findings in this cohort (Choi et al., 2018, Swanberg et al., 2021).

Notably, the open-label, single-arm nature of this pilot study design, i.e., lack of double blinding for the therapy condition plus a relapsing-remitting multiple sclerosis control group randomized into a double-blinded placebo condition, does not support definitive conclusions regarding the relationship between temporal alterations in metabolite concentration measurements and Tecfidera® therapy per se. Similarly, the abbreviated time series of control measurements (baseline and M₆ only) does not support clear modeling of a time × group interaction for the time point (M₁₂) at which a putatively therapy-group-specific rise in glutathione was observed, and truncating longitudinal comparisons between the multiple sclerosis and control cohorts to M₆ may provide a biased picture of sustained treatment effects. Extensions of this work could benefit from both placebo control and an age- and sex-matched healthy control group scanned on a schedule equivalent to that of the multiple sclerosis group. A much larger sample size would also enable more systematic analysis of potential interactions between treatment and possibly salient patient characteristics like demographic, disease duration, treatment history, and other clinical variables not currently possible for this small pilot design.

Significant differences in prefrontal total choline as well as occipital glutamine and myoinositol were observed between baseline and M₆ in healthy control participants. These observed changes in control metabolites do not support a clear assumption of cortical metabolic stability over a several-month time frame, even if the methodological stability of the ¹H MRS measurement procedures employed in this study have been explicitly validated with same-day test–retest assessment (Prinsen et al., 2017). This finding underlines the need to extend the present study design to one comparing the magnitude of difference from baseline seen in prefrontal glutathione following fumarate therapy administration with that of a matched placebo MS patient group. Similarly, decreases in occipital cortex total choline concentration were observed in the multiple sclerosis group starting at one month after baseline. In vivo cortical total choline concentration has been inconsistently associated with disorders thought to involve neuroinflammation (Chang et al., 2013); multiple sclerosis in particular has been linked with both increases and reductions in the metabolite across clinical course and tissue type (Swanberg et al., 2019). The resultant ambiguity of interpreting unidirectional changes in total choline concentration as a consequence of either therapy or disease progression in a single group undergoing therapy only further emphasizes the importance of a placebo-controlled multiple sclerosis cohort providing longitudinal context more confidently associated with disease alone.

Changes in the ¹H-MRS signal magnitude for reference metabolite creatine were observed over the course of the study. These changes are not necessarily indicative of changes in creatine concentration, a claim that cannot be made without reference to another signal of estimated concentration like water. They are also not explanatory of the rise in glutathione concentration seen from baseline to M₁₂ in the treated multiple sclerosis group, as sLASER creatine signal changes increased over time in this group and voxel, thereby proceeding in a direction opposite to that which would be expected for driving higher apparent concentrations of any metabolites using them as a reference of assumed constant value. Indeed, significant changes in creatine signal do not correspond in direction and timing with simultaneous observations of significant change in any metabolite concentration reported in this study. It is also worthy of note that no new active lesions in the voxels of interest were observed among the multiple sclerosis group participants throughout the study duration.

In addition, ¹H-MRS metabolite concentration referencing via voxel water alone carries the potentially confounding influences of different water relaxivity in multiple sclerosis versus control tissues, especially those containing lesions (Bellenberg et al., 2013, Helms, 2001, Papanikolaou et al., 2004, West et al., 2014), not to mention the more elusive possible confound of systematically differing water molarities in diseased versus control tissue. For these reasons creatine but not water-referenced metabolite concentrations were reported in the present analysis.

Glutathione is by no means independent of its biochemical environment. For example, rates of glutathione synthesis depend in part on extracellular glutamate concentrations (Frade et al., 2008). Glutamate is the major excitatory neurotransmitter in the brain, a precursor of γ-aminobutyric acid (GABA), and with cysteine and glycine one of the three amino acids comprising the glutathione tripeptide. Glutamate is closely linked to glutamine by the glutamate-glutamine cycle (Erecińska and Silver, 1990), a process that depends on the neurotransmission rate (Rothman et al., 2011). As such, the link between glutathione and other metabolites is dynamically influenced by brain function and neurotransmitter homeostasis.

Several metabolites involved in glutathione metabolism play a role in MS pathology themselves. Excess extracellular glutamate has been reported to cause calcium-mediated apoptosis in an in vitro model of MS, and a lack of oligodendrocytic glutamate transporters has been speculated as a cause for excitotoxicity (Pitt et al., 2003). Moreover, recent in vivo studies have strongly tied alterations in genes associated with glutamate metabolism to MS markers of injury (Azevedo et al., 2014, Baranzini et al., 2010). The immediate role of the inhibitory neurotransmitter GABA in MS is still unknown, but its anti-inflammatory potential has been suggested based on the concurrent observations of increased GABAergic activity and reduced inflammation in animals with experimental autoimmune encephalomyelitis (EAE), a murine model of brain inflammation (Bhat et al., 2010, Mohamed et al., 2003). ¹H-MRS has also previously demonstrated abnormally low cortical GABA in human MS patients (Cao et al., 2018, Cawley et al., 2015, Swanberg et al., 2021).

With STEAM, a clear separation of the glutamate and glutamine (plus NAA) resonances at 2.35 ppm and 2.45 ppm, respectively, was achieved. Macromolecule signals also represent a known problem for short-T_E STEAM as they are known to cause a wavy baseline and to thereby substantially complicate metabolite quantification (Behar et al., 1994, de Graaf et al., 2006). Macromolecule concentrations are potentially elevated in MS (Mader et al., 2001) and thus the risk of erroneous metabolite quantification is further enhanced for STEAM investigations of MS pathology. We therefore applied the best available MRS methods to detect the key metabolites of MS pathology and optimized them to minimize the risk of systematic errors. The use of inversion-recovery preparation for the minimization of macromolecule signals came at the cost of a substantial signal-to-noise ratio reduction. This choice was considered justified in this study to achieve the most reliable metabolite quantification. This technique is additionally limited by the macromolecule-specific spread of T₁ relaxation times, precluding the perfect cancellation of all macromolecule resonances. These influences may have contributed to measurement variability and consequent null findings for glutamate in this study, but high test–retest reproducibility for this metabolite achieved by the present methods (Prinsen et al., 2017) as well as a previous finding employing the same procedures to report in vivo cortical glutamate alterations specific to progressive and not relapsing-remitting multiple sclerosis (Swanberg et al., 2021) suggest against this interpretation.

Pioneering work in the field of non-invasive quantification of glutathione with in vivo ¹H MRS, especially at ultra-high fields, has been performed by Terpstra et al. (2003). To date, however, even the basic questions about antioxidant defense mechanisms in the human brain and the involvement of glutathione metabolism therein have not been satisfactorily answered. More importantly for the study at hand, the role of glutathione metabolism in MS pathology remains largely unknown. Our results extend previous in vivo ¹H-MRS observations of a possibly tissue-specific (Srinivasan et al., 2010) link between cortical glutathione concentrations and multiple sclerosis, particularly its progressive manifestation (Choi et al., 2018, Choi et al., 2017, Choi et al., 2011, Swanberg et al., 2021), to further suggest a role for this molecule in therapies shown to be effective against its relapsing-remitting clinical course, though this hypothesis awaits replication with a larger-scale and ideally placebo-controlled study design able to explicitly link therapy-associated changes in glutathione with measurable clinical benefits. Taken together, these findings justify further large-scale investigations of the role of glutathione and related molecules like glutamate in the evolution and treatment of not only progressive but also relapsing-remitting multiple sclerosis. Of particular import to these endeavors will be further studies employing appropriate acquisition methods like adiabatic localization, spectral editing, and dedicated macromolecule handling, as well as suitable hardware like high field strengths and correspondingly higher-order shim systems, necessary for reliably quantifying this antioxidant of interest in repeated measures over time.

CRediT authorship contribution statement

Christoph Juchem: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Supervision, Project administration. Kelley M. Swanberg: Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Visualization. Hetty Prinsen: Conceptualization, Validation, Investigation, Data curation, Writing – review & editing. Daniel Pelletier: Conceptualization, Methodology, Validation, Resources, Writing – review & editing, Supervision, Validation, Writing – review & editing, Project administration, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.