Altered in vivo brain GABA and glutamate levels are associated with multiple sclerosis central fatigue

Abstract

Purpose:

Fatigue is a common symptom in patients with multiple sclerosis (MS) with unknown pathophysiology. Dysfunction of the GABAergic/glutamatergic pathways involving inhibitory and excitatory neurotransmitters such as γ-aminobutyric acid (GABA) and glutamine + glutamate pool (Glx) have been implicated in several neurological disorders. This study is aimed to evaluate the potential role of GABA and Glx in the origin of central fatigue in relapse remitting MS (RRMS) patients.

Methods:

24 RRMS patients and 16 age- and sex-matched healthy controls (HC) were scanned using Mescher-Garwood point resolved spectroscopy (MEGA-PRESS) with a 3 T system to quantify GABA+ and Glx from prefrontal (PFC) and sensorimotor (SMC) cortices. Self-reported fatigue status was measured on all participants using the Modified Fatigue Impact Scale (MFIS).

Results:

RRMS patients had higher fatigue scores relative to HC (p ≤ 0.05). Compared to HC, Glx levels in RRMS patients were significantly decreased in SMC (p = 0.04). Significant correlations were found between fatigue scores and GABA+ (r = −0.531, p = 0.008) and Glx (r = 0.511, p = 0.018) in PFC. Physical fatigue was negatively correlated with GABA+ in SMC and PFC (r = −0.428 and −0.472 respectively, p ≤ 0.04) and positively with PFC Glx (r = 0.480, p = 0.028).

Conclusion:

The associations between fatigue and GABA + and Glx suggest that there might be dysregulation of GABAergic/glutamatergic neurotransmission in the pathophysiological mechanism of central fatigue in MS.

1. Introduction

Fatigue is a common symptom in multiple sclerosis (MS), characterised by lack of both mental and physical energy [1]. It is primarily central in nature with cognitive, physical and psychosocial elements [1]. The central origin of MS fatigue has been well documented by electrophysiological studies pointing to dysfunctions in the inhibitory intracortical connections [2], enhanced corticomotor excitability and impaired drive to the primary motor cortex [3] as possible contributing factors. Many other neuronal factors have also been proposed for the pathophysiological mechanism of this symptom including dysfunction in premotor, limbic, basal ganglia or hypothalamic region; mitochondrial dysfunction and altered central nervous system functioning caused by a disruption of the immune response and changes in metabolic levels [4]. However, the information on the exact metabolic nature of this symptom is scant.

Magnetic resonance spectroscopy (MRS) is a non-invasive tool capable of measuring the levels of brain metabolites that may potentially be involved in the MS pathophysiology. The collective evidence from several MRS studies in patients with MS fatigue showed lower N-acetylaspartate/creatine (NAA/Cr), increased myo-inositol/creatine (mI/Cr) and glutamine + glutamate/creatine (Glx/Cr) ratios in frontal, parietal, and occipital lobes as well as in hypothalamic regions of patients with severe fatigue compared to a non-fatigued group or healthy subjects [5–9]. In addition, many biochemical neuronal processes including reduced glucose metabolism in the frontal cortex and basal ganglia have been implicated in the development of fatigue in MS [10, 11] but the changes in inhibitory and excitatory neurotransmitters in patients with this symptom are currently unknown.

γ-aminobutyric acid (GABA) is the principal inhibitory neurotransmitter in human cortex responsible for the regulation of neuronal activity through GABAergic inhibition, with major impact on information processing, plasticity and network synchronization [12]. In MS, GABA plays an important role in the pathophysiological mechanism of neurodegeneration, neuroprotection and functional reorganisation [13]. There is also a suggestion that a partial loss of compensatory downregulation of cortical inhibitory mechanisms involving reduced levels of GABAergic activity is associated with MS fatigue [14]. Glutamate, on the other hand, is abundant in the brain at approximately 5–15 mmol/kg brain tissue and plays a crucial role in the cerebral metabolism by being at the crossroads of many metabolic pathways [15]. Abnormal levels of glutamate have been reported in MS patients with different disease phenotypes; increased levels in active lesions and white matter lesions and low levels in gray matter regions [16]. However, quantifying these metabolites in vivo is challenging with standard single voxel methods due to its inherently low concentration and strong overlap with adjacent higher concentration metabolites [17]. As a result, glutamate and the chemically similar glutamine are very difficult to distinguish at clinical field strength, and their composite measure Glx (Glu + Gln) is commonly reported. J-difference spectral editing sequence MEscher-GArwood point resolved spectroscopy (MEGA-PRESS) and advanced processing algorithms allow reliable and reproducible quantifications of GABA and Glx simultaneously [17,18]. Many studies utilised this technique successfully in measuring levels of these metabolites in several neurological diseases including recent MS studies [19,20]. Despite many electrophysiological studies focussing on central fatigue in MS, the relationship between in vivo GABA and Glx levels and their association with MS related central fatigue has not been explored. Given the evidence from electrophysiological studies on the role of inhibitory activity in the cortical structures, investigating the changes in these neurotransmitters levels in the prefrontal cortex or PFC (responsible for information processing) and sensorimotor cortex or SMC (primary sensory and motor control area) has the potential to enhance our understanding of their role in MS central fatigue. We hypothesised that altered levels of GABA and Glx in these anatomical locations might be associated with central fatigue in MS. The current study was therefore aimed to explore the association between fatigue scores in a homogenous relapse remitting MS (RRMS) patient group and GABA and Glx using MEGA-PRESS technique.

2. Material and methods

2.1. Participants

The inclusion criteria for patients were 1) fulfilling RRMS course according to McDonalds’ criteria, and 2) Expanded Disability Status Scale (EDSS) score between 1 and 4. Exclusion criteria include patients who were treated with GABA agonists (Baclofen), fatigue relief medications (Amantadine/Modafinil) or steroid treatment and history of recent relapse (last three months) prior to the recruitment. All patients had been undergoing disease modulating therapy with Natalizumab as part of their standard clinical care. All participants were assessed for their 1) eligibility for MRI to ensure no contraindications, 2) prior/current psychiatric or other neurological disorders and 3) any history of medications that could affect their metabolic level including supplements. Participants were also instructed to abstain from alcohol, caffeinated beverages and energy drinks. Approval to conduct this study was provided by the local health district ethics committee (approval number: 15/09/16/4.04). Informed written consent was obtained from all participants.

2.2. Clinical assessments

All patients underwent neurological examination for disability status by their treating neurologist using EDSS. MS severity score (MSSS) was also calculated. Fatigue was assessed with the modified fatigue impact scale (MFIS). The MFIS is an abbreviated version of the 40 item fatigue impact scale modified into a 21-item scale that contains three fatigue domains; overall fatigue, physical fatigue and cognitive fatigue [21]. The number of questions for each domain is as follows; cognitive (10), physical (9) and psychosocial (2). Depression and anxiety stress scale (DASS) was used to record mood levels. A neuropsychological test for cognitive performance was also performed using audio recorded cognitive screen (ARCS), which measures executive functioning/attention, memory, language, verbal fluency and visuospatial functioning [22]. Written version of Symbol digit modalities test (SDMT) was used to assess information processing speed [23]. All clinical assessments were performed prior to or immediately after on the day of their imaging session.

2.3. Imaging and spectroscopy

Structural MRI and MRS data were collected using a 3 T system (Prisma, software version: VE11C, Siemens Healthineers) equipped with a 64-channel receiver-only head coil. MRS data for GABA and Glx was acquired from two lesion free anatomical locations namely right PFC and SMC, as shown in Fig. 1. The rationale for choosing right side for voxel placement is based on evidence previously published [4,24].

Fig. 1.

MRS voxel locations on T₁-weighted images A: prefrontal cortex, B: right sensorimotor cortex using MEGA-PRESS editing sequence with TR/TE: 2000/68 ms, voxel (shaded white regions) size: (A) 18.75/ (B) 15.62mm³. The editing pulses were centered at 1.90 ppm to refocus the evolution of the 3-ppm GABA multiplet.

2.4. MRI/MRS data collection and quality assessment

Structural imaging included three-dimensional volumetric T1-weighted isotropic gradient echo (MPRAGE) with using TR/TE: 2000/3.5 ms, field of view: 25 cm², slice thickness: 1 mm, voxel size: 1 × 1 × 1 mm³ and fluid-attenuated inversion recovery (FLAIR) isotropic T2-weighted spin echo sequence with imaging parameters of TR/TE: 5000/386 ms, field of view: 25 cm², slice thickness: 1 mm, pixel size: 1 × 1 × 1 mm³ for atrophy measurements/tissue segmentation and lesion quantification, respectively. The parameters used for MEGA-PRESS were TR/TE: 2000/68 ms; voxel size: 18.75 mm³ (PFC) and 15.62 mm³ (SMC); bandwidth: 1200 Hz; averages: 128; scan duration: 9 min. An unsuppressed water reference scan was also acquired with the same parameters as MEGA-PRESS sequence with single average immediately after the editing sequence. A single-voxel non-edited short echo time (TE 30 ms) PRESS sequence with the same voxel sizes and from the same anatomical locations as MEGA-PRESS was also acquired with following parameters: TR/TE: 2000/30 ms, averages: 48, scan duration: 1.5 min, for the quantification of NAA and total creatine (creatine + phosphocreatine). All structural and MRS data collection protocols remained unchanged for all participants. The spectral quality was assessed by means of unsuppressed water peak full width at half maximum (FWHM) while fitting accuracy (fit error) were measured using standard deviation of fitting residual/amplitude of the model of both water and creatine peaks. Data with a fit error of over 15 % were not included in the statistical analysis. The estimated GABA + and Glx levels are expressed millimolar (mM). Each anatomical location required 11 min (MEGA-PRESS and PRESS) including water reference scans and manual shimming, resulting in an average overall acquisition time of about 30 min.

Since the edited GABA signal in the difference spectrum contains co-edited contributions from macromolecules and homocarnosine [16], the edited GABA signal is referred to as “GABA+” from here on.

2.5. Spectral analysis

The spectral fitting, co-registration, and segmentation of MEGA-PRESS data were performed using Gannet 3.0, a Matlab-based quantitative analysis toolkit for analysing MEGA-PRESS data [25]. The Gannet toolkit applies four different modules in the course of its processing pipeline: Gannet Load, Gannet Fit, Gannet CoRegister and Gannet Segment, to obtain tissue and CSF corrected estimates. The Gannet Load module includes zero-filling, Gaussian line broadening (3 Hz) and frequency-and-phase correction using the spectral registration algorithm [26] (Fig. 2A). The Gannet Fit module uses a single Gaussian peak to estimate the area under the edited GABA + peak at 3.0 ppm, and a double Gaussian peak to model the area under the Glx peaks at 3.74 ppm, including linear, sine, and cosine baseline terms, using a nonlinear least-squares fitting approach (Fig. 2B). Gannet CoRegister registers the MRS voxel to the T₁-weighted image to create a binary mask with the same geometric parameters as the T₁-weighted image (Fig. 2C). Finally, Gannet Segment calls statistical parametric mapping (SPM) [27] software to segment the T₁-weighted image, then uses the segmentation results to determine the tissue type fractions (gray matter, white matter and CSF) for the voxel and return CSF-corrected GABA + and Glx.

Fig. 2.

Gannet post-processing modules. A: Gannet Load module shows spectrum pre (red) and post (blue) frequency/phase corrections, B: Gannet Fit module shows modelling of the Glx and GABA + signals, ‘residual’ is the difference between the best fit ‘model’ (red) and experimental ‘data’ (blue); C: Gannet Segment shows segmented different tissue types on T1 weighted axial image on prefrontal cortex in a RRMS patient (left to right: voxel location, WM: white matter, GM: gray matter, CSF: cerebro spinal fluid).

2.6. Brain volume and T2 lesion quantification

Normalised (for each patient head size) whole brain volume was quantified using SIENAX software available on FSL package [28]. The T2 lesions (white matter lesion load (WMLL)) were calculated from FLAIR sequence which were segmented using lesion growth algorithm in the LST [29] toolbox (version 2.0.6 for SPM).

2.7. Statistical analysis

Statistical analyses were performed using SPSS Version 24.0. Normality distribution was tested and confirmed for both clinical and spectral data using Kolmogorov-Smirnov and Shapiro-Wilk tests. The group (RRMS vs HC) differences of the mean for clinical variables and metabolites were calculated using Student’s t-test. Correlations between clinical variables and metabolites were performed using Pearson correlation coefficient. Partial correlations were also performed including age, disease duration, EDSS, and depression as covariates. All data were expressed in mean ± standard deviation (Mean ± SD). Benjamini-Hochberg correction was carried out to control for the false discovery rate for multiple test comparison.

3. Results

3.1. Demographics and clinical characteristics

Twenty-four clinically definite RRMS patients aged between 23–54 years and sixteen age- and sex-matched healthy controls (HC) were enrolled into the study. Demographic characteristics and clinical details of participants are shown in Table 1. Compared with HCs, the scores for all domains of fatigue for RRMS were statistically higher (p ≤ 0.01). The RRMS patients also performed worse in cognitive assessments with ARCS and SDMT (p ≤ 0.05). RRMS patients had significantly more anxiety, but no statistical difference was observed between the groups for depression, stress or visuospatial scores.

Table 1.

Demographic details and clinical characteristics of study participants.

Parameter	Healthy control (N = 16)	RRMS (N = 24)	p value
Age	38.33 ± 9.70	39.27 ± 8.69	0.762
Disease duration	–	7.70 ± 1.04	–
MSSS	–	3.62 ± 0.54	–
EDSS	–	2.39 ± 0.36	–
MFIS	16.31 ± 10.10	31.15 ± 14.03	0.002
Physical fatigue	6.38 ± 4.70	15.30 ± 8.45	0.001
Cognitive fatigue	9.92 ± 6.15	15.85 ± 7.28	0.018
Depression	3.38 ± 3.4	5.30 ± 4.64	0.182
Anxiety	1.69 ± 1.26	5.50 ± 4.24	0.004
Stress	8.00 ± 4.39	10.94 ± 7.87	0.217
Visuospatial	101.15 ± 1.86	95.15 ± 15.77	0.108
Attention	102.00 ± 9.73	92.30 ± 15.00	0.032
ARCS	96.31 ± 13.20	84.95 ± 18.56	0.049
Total Brain volume (mm3)	1647.99 ± 22.10	1572.64 ± 77.97	0.043
Total WMLL (mm3)	–	4.9 ± 1.23	–

Values are expressed as mean ± standard deviation. Statistical significance was set at p ≤ 0.05 after multiple comparison correction.

3.2. MRS data quality (Spectral quality and fitting error)

Table 2 shows the MRS data quality for both anatomical locations and metabolites from healthy controls and RRMS group. No statistically significant differences were noted either for fit error or for spectral width between the two groups for any locations or metabolites.

Table 2.

MRS data quality represented by Fit Error and water peak FWHM.

	PFC		SMC
Fit Error (%)	GABA+	Glx	GABA+	Glx
HC	6.40 ± 1.68	6.17 ± 1.56	4.21 ± 0.64	3.50 ± 1.23
RRMS	7.74 ± 3.08	6.48 ± 2.63	4.29 ± 1.02	3.58 ± 0.73
p value	0.11	0.68	0.81	0.82
FWHM (Hz)
HC	8.80 ± 1.12		8.42 ± 0.80
RRMS	11.20 ± 1.71		11.04 ± 1.14
p value	0.83		0.99

Comparative MRS data quality between healthy subjects and patient group for both anatomical locations and metabolites. The values are expressed as mean ± SD. FWHM: full width half maximum; Hz: hertz. Statistically significant when p ≤ 0.05.

3.3. Group metabolic differences (GABA + and Glx)

Metabolic data from HC and RRMS groups are shown in Table 3. The RRMS showed significantly lower NAA/Cr ratio relative to healthy control (p ≤ 0.02) in both voxels. Compared to HC, statistically significant reduction in Glx was observed in the patient group in SMC (p = 0.04). No group differences were noted for GABA+ in either voxel. Glx/GABA+ ratios also did not show any statistical difference between the groups.

Table 3.

Selected metabolic data from PFC and SMC in HC and RRMS.

	HC	RRMS	p value	F value
Prefrontal cortex (PFC)
GABA+	1.95 ± 0.94	1.86 ± .84	0.81	0.58
Glx	6.77 ± 2.03	6.92 ± 3.48	0.84	0.04
Glx/GABA+	3.98 ± 1.40	3.72 ± 1.51	0.63	0.23
NAA/Cr^	1.68 ± 0.40	1.34 ± 0.13	0.005*	9.24
Sensorimotor cortex (SMC)
GABA+	2.05 ± 0.71	1.74 ± 0.63	0.16	2.04
Glx	6.03 ± 1.48	4.92 ± 1.60	0.04*	4.26
Glx/GABA+	3.14 ± 0.94	2.90 ± 0.48	0.41	0.73
NAA/Cr^	1.51 ± 0.18	1.36 ± 0.13	0.02*	5.52

Between group differences in GABA+, Glx and NAA/Cr obtained from voxels in PFC and SMC.

^{^}Data obtained from non-edited spectrum; all data from edited spectrum are CSF-corrected;

^*statistically significant. The values are expressed as mean ± SD.

3.4. Association between clinical variables and MRI (brain volume/lesions load)/MRS (GABA + and Glx)

The healthy controls did not show any statistically significant associations with fatigue scores for any metabolites in either voxel locations (data not shown). The statistically significant associations between clinical parameters and MRI/MRS metrics in RRMS are listed in Table 4. Cognitive function (ARCS) and information processing speed (SDMT) scores were positively correlated with GABA+, while total white matter lesion load (WMLL) has shown negative correlation with GABA + . Total fatigue score (MFIS) and physical fatigue scores showed moderate negative correlations with GABA + levels in PFC and positive correlations with PFC-Glx (Fig. 3A&C). GABA + in SMC showed negative correlation with physical fatigue (Fig. 3B). Total intracranial brain volume and T2 lesion load did not show any correlation with any of the fatigue scores, but EDSS was positively correlated with physical fatigue. Similarly, ARCS and SDMT also did not show any correlation with fatigue scores but both were positively correlated with PFC GABA+. These correlations were shown both on Pearson and partial correlation tests after adjusting for age, EDSS, disease duration and depression. Cognitive fatigue did not show any correlation with GABA + but negatively correlated with SMC-NAA/Cr.

Table 4.

Correlations between clinical variables and selected MRS parameters in RRMS.

Clinical variables	Metabolites	PFC		SMC
		r	p value	r	p value
ARCS	GABA+	0.524	0.037
SDMT	GABA+	0.530	0.035
MFIS	GABA+	−0.531	0.008
	Glx	0.511	0.018
Physical fatigue	GABA+	−0.472	0.020	−0.428	0.037
	Glx	0.480	0.028
Cognitive fatigue	NAA/Cr			−0.446	0.029

Only statistically significant correlations present on both Pearson and partial correlations are shown. p values shown are corrected for multiple comparisons.

Fig. 3.

Scatterplot charts show association between fatigue scores and neurometabolites (GABA + and Glx) in prefrontal cortex (PFC) and sensorimotor cortices (SMC) in healthy controls (blue square and line) and RRMS group (red circle and line) A & B: The RRMS group showed statistically significant negative correlations between MFIS and physical fatigue and PFC-GABA + level and SMC-GABA+, respectively. C: Glx has shown statistically significant positive correlation with MFIS in PFC of RRMS group. These correlations were shown both on Pearson and partial correlation tests after adjusting for age, EDSS, disease duration and depression.

4. Discussion

The highlight of this study is that levels of GABA + and Glx in RRMS patients are significantly correlated with fatigue scores which provides the first in vivo evidence of the involvement of inhibitory/excitatory neurotransmitters in MS-related central fatigue. Fatigue in MS is not only multidimensional and complex but its pathophysiology is mostly unknown. Several mechanisms for the development of fatigue have been proposed including a chronic imbalance of inflammatory markers such as proinflammatory cytokines [30]. Given the effect of GABA on the immune system through modulation of proinflammatory cytokine and chemokine production [31], it is conceivable that dysregulated GABA homeostasis may have a bearing on fatigue development. Adaptive compensatory mechanism for maintaining normal function has been reported in MS patients with central fatigue [4]. The associations in our study may suggest GABA transporter 2 (GAT2) overexpression and high immunoreactivity in MS [31] or even decreased inhibitory innervation of cortical neurons leading to higher energy demands [20,32]. In addition, we observed decreased SMC-GABA + levels which is in line with the findings of Cawley et al. who found similar results (albeit in secondary progressive MS), potentially indicating reduced GABA receptor expression and decreased density of inhibitory interneurons [20]. Converging evidence from advanced MRI studies suggests the involvement of PFC in the development of fatigue in MS [4]. Functional MRI data report greater activation in PFC and SMC in fatigued patients compared to non-fatigued MS patients indicating the role of attentional and motor planning circuits in the development of central fatigue in MS [33]. Alteration and disruption of pathways/circuits within the frontal lobes mainly PFC [34] and between frontal lobes and other brain structures including striatal, limbic, pyramidal, thalamic and occipital regions in fatigued MS patients have also been suggested [4]. All these findings provide evidence of the importance of these anatomical locations in the pathogenesis of central fatigue in MS. The observed association between GABA and fatigue levels may reflect the involvement of inhibitory cortical connections in MS fatigue.

We also noticed increased levels of Glx concentration in RRMS in PFC with a positive correlation with MFIS. Glx plays a critical role in oxidative energy supply to neurons/astrocytes and production of GABA [29]. Excessive amount of Glx can lead to neurotoxicity, degeneration and dysfunction in the glutmatergic pathway [35] in addition to abnormal neuronal signalling, glutamate-induced activation of T cells and release of proinflammatory cytokines, all of which have detrimental effects in patients with MS [36]. Inflammatory cytokines are known to block, release and reuptake of glutamate by astrocytes via several pathways [37] – as reported in chronic fatigue syndrome [38]. Also, altered metabolic flow from glutamate to succinate of the tricarboxylic acid (TCA) cycle via GABA has been suggested in chronic fatigue syndrome [39]. Although the exact mechanism of fatigue development in MS is uncertain from the findings of this study, it can be theorised that dysregulation either in the synthesis of GABA from glutamate (downregulation of GAD65 and GAD67) or upregulation of GAT2 in the GABAergic pathway of the TCA cycle (Fig. 4) may have a role in MS related fatigue. In stark contrast, we found significantly lower levels of SMC-Glx level in patients. Previous studies with similar findings suggest that reduced glutamate levels in cortical areas might explain impaired glutamatergic activity due to myelin damage or reduced axo-myelinic communication in MS [40–42]. Interpretation of this change in Glx at a cellular level is however challenging because glutamate-mediated processes in MS are complex and likely not detectable with in vivo MRS [42]. Nonetheless, neuronal metabolic dysfunction or impaired glutamate homeostasis in cortical areas of patients with MS could be hypothesised.

Fig. 4.

Glutamatergic/GABAergic pathway. AT: aminotransferase; Gln: glutamine; Glu: glutamate; GABA: Ɣ-aminobutyric acid; α-KG: α-keto-glutarate; EAAT: excitatory amino acid transporter; GAD: glutamate decarboxylase; GAT: GABA Transporter; GS: glutamine synthetase; PAG: phosphate activated glutaminase; SNAT: Sodium-coupled neutral amino acid transporter; TCA: tricarboxylic acid.

As evidenced in other studies [6–8], the NAA/Cr ratio in our patient group was significantly reduced compared to HC. Decreased NAA/Cr has been well established as neural correlates for disease progression, cognitive deficits and disability due to ongoing diffuse axonal loss [43]. However, the association between NAA/Cr and central fatigue is rather contradictory as those who found correlation hypothesized fatigue may be a consequence of diffuse axonal dysfunction [7]. Others who did not find a link suggest changes in NAA/Cr could be driven by either increase in total creatine [9] or disease-related cycling of NAA and glutamate [5]. Add to the complexity, the variability in sample size, MRS method used, voxel location and degree of decrease in NAA/Cr ratio between regions which was found to be higher in parietal regions than frontal/occipital regions [44] may also have played a role in the diverging results. We only found a negative association between SMC-NAA/Cr and cognitive fatigue in this study. Diffuse axonal degeneration could potentially explain the cognitive deficits we observed in our patients and lead to a secondary fatigue. Nonetheless, the association of NAA/Cr ratio with cognitive fatigue in this study is rather moderate compared to the relationship between GABA and fatigue. This points to the fact axonal loss alone does not explain the development of fatigue but some other mechanisms, possibly dysfunction along thalamo-frontal neuronal circuits, are involved as well.

However, the lack of significant group difference in GABA measurements in our study can be explained by the confounding variability in GABA levels in different brain regions as evidenced in previous studies [45] despite using the same measurement technique such as MEGA-PRESS. One must bear in mind that the synthesis of GABA is a dynamic process [15] that is influenced by various factors such as age/gender [46], tissue composition and menstrual cycle [47], use of nicotine, alcohol, caffeine, and even childhood trauma [48] in addition to disease related changes in patients. Although we endeavoured to account for all above factors except menstrual cycle in our study as well as applying a rigorous statistical method, our study is limited by a small sample size which could explain the loss of significance between the two groups. In line with our findings, Cao et al. also could not establish a significant change in PFC-GABA levels between MS patients and HC [19] which they report due to diminished inhibitory influence of GABA on neural circuits as evidenced in pain perception studies [49]. Similarly, another study found no significant group difference for SMC-GABA levels but observed a strong correlation with physical disability score in MS patients [50].

Our results further demonstrate the link between GABA + and information processing as well as cognitive function. Significant positive associations were observed between PFC-GABA + and SDMT/ARCS (tests for information processing and cognitive function). Yoong et al. established the importance of GABA levels in dorsilateral prefrontal cortex in working memory, suggesting that the amount of GABA is a critical component in working memory capacity [51]. The correlation between PFC-GABA + level and cognitive function in our study was also in concordance with another study indicating cognitive function is sensitive to changes in GABA levels in the frontal cortex [52]. However, Nantes et al. noticed elevated SMC-GABA levels in MS patients are associated with better SDMT scores [41]. Although we could not establish an association with SMC-GABA + and SDMT in our RRMS patients, the SMC-GABA + levels were relatively lower. A recent study by Cao et al. showed a similar connection of lower levels of GABA + and worse cognitive performance in the RRMS group in the hippocampus and posterior cingulate cortex [19]. In another study involving secondary progressive MS, no associations between PFC-GABA and cognition could be identified [20]. These discrepancies in the level of GABA + and its associations could be due to different mechanisms in different disease subtypes, methodological and/or regional variations. Our RRMS patients had worse cognitive performance in particular poor information processing compared to healthy controls and there was a negative correlation between cognitive fatigue and lower SDMT score as well as a positive correlation between lower ARCS and decreased level of PFC-GABA + . This implies our RRMS patients struggled with the complex mental task due to cognitive fatigue. Despite a low mean EDSS of 2.39 in our patient group, we observed a positive correlation between EDSS and physical fatigue. The association between physical disability and fatigue is a matter of controversy as studies report conflicting results, some with strong correlation while others found no relationship at all [53]. In line with previous studies [54,55], we also found no correlation between total brain white matter lesion load or total intracranial volume with any of the fatigue scores underlining the fact that metabolic assessment is more senisitive than volumetric.

4.1. Strengths/Limitations/Recommendations

Strengths of this study include homogenous low disability in a predominantly female cohort of stable RRMS with very mild depression. Although our findings are preliminary, this pioneering study forms a foundation for much more considerable research into the pathophysiological mechanism of central fatigue in MS. We also adopted a streamlined post processing spectral analysis method to address inherent limitations with spectral data analysis which enabled our data quality to be superior. Additionally, we used MFIS, a much reproducible and reliable tool to measure different domains of MS fatigue. Finally, the statistical approach used in this study was robust and subject to false discovery correction to ensure our findings are valid.

There are however limitations in our study mainly relatively small sample size which we tried to overcome by selecting a homogenous group of young female patients (mean age 39 years and only one male) who were stable on the same medication (Natalizumab). However, small sample size/large standard error may have inflated correlation coefficients observed in this study and we suggest our results should be interpreted with caution. Although we tried to adjust for most factors, we have not accounted for hormonal changes that occurred in our patients. Changes in GABA + concentration have been reported during the menstrual cycle and individual variations [48] may have diluted our results. Although the editing of GABA + with MEGA-PRESS is feasible, there may be significant macromolecule (almost 40 %) co-edited contaminations in our GABA + estimate [15]. Similarly, some signal contamination from glutathione is also present in Glx quantification [56]. The interpretation of MRS detected GABA + concentration is complex and it is still an issue of ongoing debate whether MRS observed GABA could be used as an index of GABAergic neurotransmission rather than simply being a marker of GABAergic interneuron cell density [18]. Despite the exclusion of patients with GABA agonists in our study, it is important to note that all our patients were treated with the same disease modulating therapy (Natalizumab), and the effects of this therapy on GABA + and Glx are not fully known. Finally, our voxel size is limited to two anatomical locations as regional variation in metabolic concentration has previously been reported [48] and as such we are unable to generalise the changes for the whole brain. Fatigue in MS fluctuates significantly over time and hence it is important to enhance our understanding its dynamic nature to develop targeted therapies [57]. Longitudinal trials are therefore vital to track the behaviour of metabolic concentrations. Longitudinal studies further allow prediction of future changes in fatigue using the baseline information as well as the ability to follow up interventions/therapy [58].

5. Conclusion

In conclusion, the findings of our study support of altered level of GABA + and Glx to have a neurochemical association with central fatigue in MS. These findings may suggest a neurochemical link for impaired modulation of neurochemical activity and/or dysregulation of GABAergic/Glutamatergic neurotransmission in the mechanisms of mediating central fatigue. Studies involving larger sample size and heterogeneous patient cohorts are required to further confirm these findings.

Abbreviations:

ARCS

audio recorded cognitive screen

EDSS

expanded disability status scale

GABA

γ-aminobutyric acid

GABA+

GABA & co-edited contributions from macromolecules and homocarnosine

Glx

glutamine/glutamate

HC

healthy control

MRS

magnetic resonance spectroscopy

MEGA-PRESS

Mescher-Garwood point resolved spectroscopy

MFIS

modified fatigue impact scale

MSSS

multiple sclerosis severity score

NAA

N-acetylaspartate

PFC

prefrontal cortex

RRMS

relapse remitting multiple sclerosis

SDMT

symbol digit modalities test

SMC

sensorimotor cortex

WMLL

white matter lesion load