An Imbalance Between Excitatory and Inhibitory Neurotransmitters in Amyotrophic Lateral Sclerosis Revealed by Use of 3-T Proton Magnetic Resonance Spectroscopy

Importance  A lack of neuroinhibitory function may result in unopposed excitotoxic neuronal damage in amyotrophic lateral sclerosis (ALS).

Objective  To determine whether there are reductions in γ-aminobutyric acid (GABA) levels and elevations in glutamate-glutamine (Glx) levels in selected brain regions of patients with ALS by use of proton magnetic resonance spectroscopy.

Design  Case-control study using short echo time and GABA-edited proton magnetic resonance spectroscopy at 3 T with regions of interest in the left motor cortex, left subcortical white matter, and pons; data analyzed using logistic regression, t tests, and Pearson correlations; and post hoc analyses performed to investigate differences between riluzole-naive and riluzole-treated patients with ALS.

Setting  Tertiary referral center.

Participants  Twenty-nine patients with ALS and 30 age- and sex-matched healthy controls.

Exposure  Fifteen patients were taking 50 mg of riluzole twice a day as part of their routine clinical care for ALS.

Main Outcomes and Measures  Levels of GABA, Glx, choline (a marker of cell membrane turnover), creatine (a marker of energy metabolism), myo-inositol (a marker of glial cells), and N-acetylaspartate (a marker of neuronal integrity).

Results  Patients with ALS had significantly lower levels of GABA in the motor cortex than did healthy controls (P < .01). Patients with ALS also had significantly lower levels of N-acetylaspartate in the motor cortex (P < .01), subcortical white matter (P < .05), and pons (P < .01) and higher levels of myo-inositol in the motor cortex (P < .001) and subcortical white matter (P < .01) than did healthy controls. Riluzole-naive patients with ALS had higher levels of Glx than did riluzole-treated patients with ALS (P < .05 for pons and motor cortex) and healthy controls (P < .05 for pons and motor cortex). Riluzole-naive patients with ALS had higher levels of creatine in the motor cortex (P < .001 for both comparisons) and subcortical white matter (P ≤ .05 for both comparisons) than did riluzole-treated patients with ALS and healthy controls. Riluzole-naive patients with ALS had higher levels of N-acetylaspartate in the motor cortex than did riluzole-treated patients with ALS (P < .01).

Conclusions and Relevance  There are reduced levels of GABA in the motor cortex of patients with ALS. There are elevated levels of Glx in riluzole-naive patients with ALS compared with riluzole-treated patients with ALS and healthy controls. These results point to an imbalance between excitatory and inhibitory neurotransmitters as being important in the pathogenesis of ALS and an antiglutamatergic basis for the effects of riluzole, although additional research efforts are needed.

Amyotrophic lateral sclerosis (ALS) is a progressive degenerative motor neuron disease involving the motor cortex, corticospinal tract, brainstem, and spinal anterior horn neurons.¹ The disease is uniformly fatal, although the clinical presentation and course are heterogeneous, with median survival times between 2 and 4 years.² Patients present most commonly with combined upper motor neuron (UMN) and lower motor neuron (LMN) features, although, earlier in the disease course, only UMN or LMN signs may be present. Although electromyography confirms LMN signs in ALS, UMN signs are solely assessed on clinical grounds that can delay diagnosis. This is due in part to the fact that UMN signs in ALS can be masked by LMN weakness and secondary hyporeflexia rather than hyperreflexia.³ Riluzole, the only medication for ALS approved by the US Food and Drug Administration, has limited efficacy, extending average life expectancy by only 3 to 6 months.^4,5 Riluzole is postulated to modulate excitatory neurotransmission, although the exact in vivo pharmacologic actions are not well understood.⁶

To further the understanding of the disease process and perhaps to diagnose the disease at an earlier stage, advanced magnetic resonance imaging techniques have been applied to the study of ALS. Proton magnetic resonance spectroscopy (¹H-MRS) is one such imaging technique that has been used in a number of studies to investigate ALS.⁷ The techniques of ¹H-MRS require the placement of a voxel or region of interest to specify a volume of brain tissue in which to measure brain metabolites. Conventional in vivo ¹H-MRS at 3 T can quantify various brain metabolites, including N-acetylaspartate (NAA; a marker of neuronal integrity), choline (Cho; a marker of cell membrane turnover), creatine (Cr; a marker of energy metabolism), myo-inositol (mI; a marker of glial cells), and glutamate-glutamine (Glx).⁸ γ-Aminobutyric acid (GABA), the major inhibitory neurotransmitter, is difficult to quantify using conventional ¹H-MRS but can be measured using spectral-editing techniques.^9,10 There is increasing evidence that reduced inhibitory function may play an important role in the pathogenesis of ALS.¹¹ Given our limited prior knowledge of in vivo GABA changes in ALS, direct interrogation of GABA may lead to a new understanding of this complex disease and may provide opportunities for the development of new disease-modifying treatments.

The most common finding reported in ¹H-MRS studies of ALS is reduced NAA in the motor cortex, which is generally interpreted as neuronal loss.¹² Although riluzole is thought to modulate excitatory neurotransmission, only 2 published ¹H-MRS studies^13,14 have explored the effect of riluzole on brain metabolites and measured only Cho, Cr, and NAA. We recently published the first ¹H-MRS study¹⁵ reporting a decrease in GABA motor cortex levels in a small ALS cohort (n = 10), which suggested reductions in inhibitory neurotransmission. However, our prior study¹⁵ did not consider the excitatory contribution of Glx. The aim of the present study was to study both inhibitory neurotransmission, as measured by GABA, and excitatory neurotransmission, as measured by Glx. Our hypothesis is that there is an imbalance between excitatory and inhibitory neurotransmitters in ALS. The GABA and Glx levels were measured in the left motor cortex and left subcortical white matter, and Glx levels were measured in the pons as well. A secondary aim was to determine differences between brain metabolite profiles of riluzole-treated vs riluzole treatment-naive patients with ALS.

Twenty-nine patients with ALS were recruited from our institution’s ALS clinic. Thirty age- and sex-matched healthy controls were also recruited. The patients with ALS met the El Escorial Criteria for probable (n = 15), probable laboratory-supported (n = 10), or definite (n = 4) ALS.¹⁶ Participants were excluded if they had a history of central nervous system infection, head injury, or cerebrovascular disease; were active substance abusers; or had a contraindication for magnetic resonance imaging. All participants were right-handed. Our institutional review board approved all study protocols, and informed consent was obtained from all participants. The UMN scores were graded by combining the Ashworth Spasticity Scale (range, 0-8)¹⁷ with the presence of pathological reflexes (range, 0-24)^18,19 and a scale for measuring the pseudobulbar affect (range, 0-1),²⁰ with a total scale ranging from 0 to 33 (a higher score reflecting higher disease burden). Participant characteristics are presented in Table 1.

Nineteen participants (10 patients with ALS and 9 healthy controls) were imaged on a Philips Achieva 3T system using an 8-channel receive head coil; the GABA results from this cohort have been previously published.¹⁵ In addition, 40 participants (19 patients with ALS and 21 healthy controls) were imaged on a Philips Ingenia 3T system using a 15-channel receive head coil. T1-weighted 3-dimensional magnetization-prepared rapid acquisition gradient echo images were used to place 3.0 × 3.0 × 2.0-cm voxels in the left motor cortex and left subcortical white matter located caudally to the motor cortex for single-voxel point-resolved spectroscopy (PRESS) and Mescher-Garwood PRESS (MEGA-PRESS) data acquisitions.⁹ A 1.4 × 1.4 × 1.8-cm voxel was placed centrally in the pons for PRESS data acquisition (Figure 1).

The PRESS spectra (with a repetition time [TR] of 2000 milliseconds and an echo time [TE] of 35 milliseconds) were acquired using VAPOR (variable power and optimized relaxation delays) water suppression: 32 averages were performed for the motor cortex and subcortical white matter voxels, and 96 averages were performed for the pons voxels. Conventional PRESS data were analyzed using LCModel version 6.1-4A.²¹ Metabolite concentrations from LCModel were only used for statistical analysis if the Cramér-Rao lower bounds were less than 20% for the motor cortex and subcortical white matter voxels and less than 25% for the pons voxel. Cerebral spinal fluid correction was performed for each voxel using magnetization-prepared rapid acquisition gradient echo images and Statistical Parametric Mapping version 5 (Wellcome Trust Centre for Neuroimaging).

The MEGA-PRESS experiment for editing GABA was performed with the following parameters: TE = 68 milliseconds (TE₁ = 15 milliseconds and TE₂ = 53 milliseconds); TR = 1.8 seconds; 256 averages; and frequency-selective editing pulses (14 milliseconds) applied at 1.9 ppm (on) and 7.46 ppm (off). Slice-selective refocusing was performed using amplitude-modulated pulse “GTST1203” (length, 7 milliseconds; bandwidth, 1.2 kHz); MEGA-PRESS was analyzed using in-house postprocessing software in Matlab 2012a (Mathworks) with Gaussian curve fitting to the GABA and inverted NAA peaks. The GABA levels were expressed relative to the NAA signal in the edited spectra.²² The GABA:NAA ratio was then multiplied by the NAA concentration determined from LCModel analysis of a short-TE PRESS spectrum of the same voxel to provide an estimate of GABA concentration. Note that the metabolite concentration is in international units (IU) because various factors are not corrected, including editing efficiency and relaxation times.

Logistic regression analyses were performed between disease status and individual metabolites using scanner type (Achieva or Ingenia) as a covariate to determine if there were significant differences between the Philips Achieva 3T and Ingenia 3T systems. Two-tailed independent-sample t tests were performed to determine differences in brain metabolites between patients with ALS and healthy controls. Pearson correlations were performed for associations between brain metabolites and clinical status (using UMN scores, disease duration, the revised ALS Functional Rating Scale [ALSFRS-R], and the rate of disease progression [defined as (48 − ALSFRS-R score at evaluation)/disease duration from symptom onset to evaluation]). A subset analysis was also performed to compare riluzole-treated vs riluzole-naive patients with ALS. We also compared riluzole-naive patients with ALS vs healthy controls. Stata version 11 (StataCorp) was used for statistical analysis. The significance threshold was set a priori at P = .05.

For the conventional PRESS spectra, 2 patients with ALS had an inadequate signal to noise ratio from the pons voxels and were excluded, and the pons Glx value from 1 patient with ALS was excluded owing to a high Cramér-Rao bound. For the MEGA-PRESS spectra, the GABA spectra in the motor cortex of 2 patients with ALS and 1 healthy control and in the subcortical white matter of 1 patient with ALS had an inadequate signal to noise ratio. Four of the patients with ALS were unable to complete the entire imaging protocol and did not undergo ¹H-MRS of the subcortical white matter. The logistic regression analysis showed no significant effects of scanner type (z > 0.05) for all of the metabolites, which indicates that the data from the cohorts scanned on the different machines are comparable.

As shown in Figure 2, patients with ALS demonstrated significantly lower levels of GABA in the motor cortex than did healthy controls (P = .002). There were no significant differences in GABA levels in the left subcortical white matter between patients with ALS and healthy controls (P = .30). There was a significant correlation between GABA level in the motor cortex and disease duration (r = −0.39; P = .05). The patients with ALS who had the 5 lowest GABA levels had a disease duration (33 months) that was almost double that of the patients with ALS who had the 5 highest GABA levels (18 months), although the ALSFRS-R and UMN scores were close (ie, the ALSFRS-R score was 5 points lower, and the UMN score was 2 points higher, for the lowest GABA levels).

Results are summarized in Table 2. The levels of NAA in the motor cortex (P = .008), subcortical white matter (P = .02), and pons (P = .003) were significantly lower in patients with ALS than in healthy controls. The levels of mI in the motor cortex (P < .001) and subcortical white matter (P = .002) were significantly higher in patients with ALS than in healthy controls. There were significant correlations between NAA level in the motor cortex and ALSFRS-R score (r = 0.39; P < .05), between mI level in the subcortical white matter and disease duration (r = 0.43; P < .05), and between Glx level in the pons and UMN score (r = −0.63; P < .001). There was a significant correlation between NAA and GABA levels (r = 0.57; P = .002). There were no significant correlations between GABA and Glx levels within the same voxel location or between voxel locations.

Subgroup characteristics are presented in Table 3. As seen in Figure 3, there were significantly higher levels of Cr, Glx, and NAA in the motor cortex of riluzole-naive patients with ALS compared with the riluzole-treated patients with ALS. There were also significantly higher levels of Glx in the pons (P = .01) and higher levels of Cr in the subcortical white matter (P = .05) of the riluzole-naive patients with ALS compared with the riluzole-treated patients with ALS. There were no significant differences in the levels of GABA or other metabolites between the 2 subgroups. The levels of Glx in the motor cortex (P = .03) and the pons (P = .04) were significantly higher for riluzole-naive patients with ALS (mean [SD] levels of 6.49 [1.52] IU and 8.28 [2.53] IU, respectively) than for healthy controls (mean [SD] levels of 5.62 [1.00] IU and 6.83 [1.77] IU, respectively). The levels of Cr in the motor cortex (P < .001) and the subcortical white matter (P = .03) were significantly higher for riluzole-naive patients with ALS (mean [SD] levels of 5.46 [0.58] IU and 4.47 [0.30] IU, respectively) than for healthy controls (mean [SD] levels of 4.93 [0.33] IU and 4.21 [0.35] IU, respectively).

Our study demonstrates reductions of GABA levels in the motor cortex of patients with ALS compared with healthy controls. In addition, it was found that Glx levels were elevated in riluzole-naive patients with ALS compared with riluzole-treated patients with ALS and healthy controls. The inclusion of an additional 19 patients with ALS augments the significance of our prior study¹⁵ that found lower levels of GABA in the motor cortex of 10 patients with ALS. To our knowledge, this is the first report of in vivo measurements of both GABA and Glx in patients with ALS, thereby investigating both the GABAergic (inhibitory) and the glutamatergic (excitatory) neurotransmitter systems in the same patients.

The cause of ALS remains elusive, although excitotoxic neuronal injury is thought to play an important role, which is primarily mediated through glutamate toxicity.^2,23 There is increasing evidence that an “interneuronopathy” may be a central player in the pathophysiology of ALS.²⁴ Interneuronopathy is the hypothesis that inhibitory or GABAergic dysfunction results in relatively unopposed excitotoxic neuronal damage. Neuroimaging, animal, histochemical, genetic, and clinical studies support the role of an interneuronopathy in ALS.¹¹ An ALS functional magnetic resonance imaging study²⁵ demonstrated increased functional connectivity in ALS, suggesting a loss of inhibitory neuronal tone. An animal model of ALS demonstrated that cortical excitability is explained by reduced GABAergic inhibition.²⁶ In addition, human histochemical and positron emission tomography studies have implicated GABA receptor alterations in the motor cortex of patients with ALS.^27,28

A significant barrier to further supporting the interneuronal hypothesis has been the challenge of directly measuring in vivo GABA concentrations. γ-Aminobutyric acid has chemical shift overlap with other brain metabolites such as Cr, Glx, and NAA, which are present in brain tissue at higher concentrations than GABA.²⁹ The ¹H-MRS editing techniques allow for improved differentiation of metabolites and quantification of GABA by editing out the overlapping Cr peak at 3.0 ppm.¹⁰ The current result of decreased levels of GABA in the motor cortex provides further in vivo evidence of reduced inhibitory function in the pathophysiology of ALS and suggests the possibility of the development of new ALS disease-modifying treatments aimed at increasing inhibitory neuronal tone.

Higher levels of Glx were also found in the motor cortex and pons of riluzole-naive patients compared with riluzole-treated patients and healthy controls. Glutamate and glutamine are difficult to resolve independently at field strengths of 3 T or lower, and are usually reported as a combined measure (ie, Glx); however, in a normal brain, glutamate is the larger contributor to this peak by about a 4:1 ratio.³⁰ Owing to concerns of a potential relationship between NAA and Glx, we confirmed that there was not a significant correlation between the NAA levels and the Glx levels for the overall group or the treatment subgroups. Riluzole is thought to act on both the glutamatergic and GABAergic systems.^31,32 Riluzole has been shown to decrease glutamatergic neurotransmission by acting as an antagonist of presynaptic N-methyl-d-aspartate and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid glutamate receptors, as well as by increasing glutamate transporter uptake.^33-35 Riluzole has been shown to increase GABA levels in cell cultures, although the levels required for GABA modulation are thought to be higher than those required for glutamatergic inhibition, which may explain the lack of effect of riluzole on the GABA levels seen in the present study.^36,37 Reduced motor cortex excitability has been demonstrated after riluzole administration in healthy subjects, as well as partial normalization of increased cortical excitability in patients with ALS, which is thought to be mediated through glutamatergic rather than GABAergic interactions.^38,39

A moderate negative correlation between GABA levels and disease duration was found. The patients with ALS who had the lowest GABA levels had longer disease durations than did the patients with ALS who had the highest GABA levels. It may be that patients with ALS who have a more rapid course have a reflexive increase in inhibitory tone relative to those with a more prolonged disease course. There may be a release of GABA from rapidly denervated interneurons resulting in higher relative levels in patients with ALS who had a short disease duration. Interestingly, both Filippini et al⁴⁰ and Iwata et al⁴¹ found in their cross-sectional studies that fractional anisotropy, a measurement of white matter integrity, was paradoxically higher in patients with ALS who had the longest disease duration. Fractional anisotropy is a different advanced neuroimaging metric, but, taken together, these findings suggest that long-term survivors have different pathophysiologic processes. Although there was no significant direct correlation between GABA and Glx levels, it is possible that there may be a relative GABA threshold for each respective patient, allowing for the excitotoxic effects of glutamate rather than a linear relationship between GABA and glutamate. It is interesting to note that there was a significant correlation between NAA and GABA levels, which suggests an association between neuronal integrity and neuronal inhibitory tone. A longitudinal imaging trial is required to address these issues.

The riluzole-treated patients with ALS had lower levels of Cr than both the riluzole-naive patients with ALS and the healthy controls. It has been proposed that an energy-depleted state may be an inciting factor in ALS³¹ and that riluzole may reduce neuronal cellular energy demand.⁴² A potential confounding factor to these findings is that the riluzole-treated patients with ALS had lower levels of NAA than did the riluzole-naive patients with ALS. There was a moderate correlation between NAA and Cr levels in the motor cortex (r = 0.49; P = .01) for the 29 patients with ALS. Two prior studies^13,14 have reported a longitudinal increase in the NAA:Cr ratio of patients with ALS after short periods (≤3 weeks) of riluzole treatment; however, these 2 studies^13,14 did not report absolute concentrations of the NAA and Cr metabolites, which makes direct comparison difficult. However, overall, these results do suggest the potential of ¹H-MRS measurements of GABA and Glx as surrogate markers of disease progression and treatment response, which might be useful in future pharmacological trials.

As in our study, almost all prior ¹H-MRS studies have reported decreases in levels of NAA or NAA:CR ratios, particularly in the motor cortex. The finding of elevated mI levels in the motor cortex and subcortical white matter is thought to represent increased numbers of glial cells.⁸ Although mI has not been investigated to the same extent as NAA, prior studies^43-45 have reported mI (or mI:Cr ratio) elevations in the motor cortex. Interestingly, only a few studies have measured Glx, with 2 studies reporting elevations in Glx as reported by increased Glx:Cr ratios in the medulla in the study by Pioro et al⁴⁶ and increased Glx:Cr ratios in the motor cortex in the study by Han and Ma.⁴⁷ Of the prior studies that measured Glx, it is important to note that riluzole status, which is a potential mediating factor (as the present study would indicate), was either not mentioned^43,46-48 or unclear.⁴⁴

As reported before, the limitations of ¹H-MRS (including the MEGA-PRESS technique) include its inability to differentiate between the intracellular and extracellular contribution of the metabolites, its potential macromolecular component contributions, and the relatively large voxel sizes required. Our study includes the GABA data of 10 patients with ALS and 9 healthy controls from previously published results. We did not perform a statistical correction for multiple comparisons, which can result in type I errors. As such, subsequent studies are needed to confirm our findings. We were not able to implement the MEGA-PRESS sequence to measure GABA levels in the pons given the relatively small volume. The riluzole-treated group of patients with ALS had, on average, a lower ALSFRS-R score than did the riluzole-naive group of patients with ALS, indicating greater disability, although the UMN disease burden was not significantly different between these 2 groups. Direct causality between riluzole treatment and effect on brain metabolites cannot be established given the cross-sectional nature of our study. A longitudinal imaging trial enrolling patients at initial diagnosis would be required to better establish the response of brain metabolites to treatment, as well as to probe ALS central nervous system changes, including the direct relationship between GABA and Glx over time.

In conclusion, reductions in GABA levels in the motor cortex of patients with ALS were observed, as well as elevations of Glx in riluzole-naive patients with ALS compared with riluzole-treated patients with ALS and healthy controls. The results support the hypothesis that an imbalance between excitatory and inhibitory neurotransmitters is an important factor in the pathogenesis of ALS, as well as the antiglutamatergic basis for the effects of riluzole. These findings also support the potential of ¹H-MRS to establish Glx and GABA measurements as clinically relevant markers of disease, although additional research efforts are needed to better understand the findings reported herein.

Accepted for Publication: January 25, 2013.

Corresponding Author: Bradley R. Foerster, MD, Department of Radiology, University of Michigan, 1500 E Medical Center Dr, UH B2 A205H, Ann Arbor, MI 48109-5030 (compfun@umich.edu).

Published Online: June 24, 2013. doi:10.1001/jamaneurol.2013.234.

Author Contributions:Study concept and design: Foerster, Pomper, Mohamed, Carlos, Barker, Feldman.

Acquisition of data: Foerster, Callaghan, Petrou.

Analysis and interpretation of data: Foerster, Pomper, Edden, Mohamed, Welsh, Carlos, Feldman.

Drafting of the manuscript: Foerster.

Critical revision of the manuscript for important intellectual content: Pomper, Callaghan, Petrou, Edden, Mohamed, Welsh, Carlos, Barker, Feldman.

Statistical analysis: Foerster.

Obtained funding: Foerster, Mohamed, Feldman.

Administrative, technical, and material support: Edden, Mohamed, Welsh.

Study supervision: Foerster, Pomper, Feldman.

Conflict of Interest Disclosure: Dr Carlos has served as a consultant for Philips Healthcare.

Funding/Support: This study funded by the A. Alfred Taubman Medical Research Institute.

Additional Contributions: We thank Andrea Smith, MS, for her assistance in the coordination of the study.