The Role of GABA in Primary Insomnia

The authors have indicated no financial conflicts of interest. There are several lines of evidence that suggest γ-aminobutyric acid (GABA), the primary inhibitory neurotransmitter in the central nervous system (CNS), plays a role in the etiology of chronic insomnia. In this issue of SLEEP, Morgan et al.1 report a 12% increase in occipital GABA, measured using proton magnetic resonance spectroscopy (1H-MRS), in primary insomnia (PI) subjects relative to healthy controls. Moreover, in both groups, GABA levels correlated negatively with wake after sleep onset measured by ambulatory polysomnography. The authors hypothesized that increased occipital GABA in PI may reflect an adaptive compensatory response to chronic hyperarousal in the disorder.1 At first glance, the finding of increased GABA in PI is counterintuitive, given the role GABAergic nuclei play in sleep initiation and maintenance,2,3 the promotion of sleep by benzodiazepine receptor agonists via enhanced receptor affinity for GABA,4 and existing data regarding hyperarousal, brain activity, and GABA. In healthy subjects, 1H-MRS-derived GABA correlates inversely with cortical activity (measured using functional MRI).5 In PI, data from positron emission tomography (PET),6 transcranial magnetic stimulation,7 and evoked-relatedpotential investigations8,9 support CNS hyperarousal in the disorder, which would be theoretically characterized by reduced, rather than increased, cortical GABA. Two previous studies from our group have utilized 1H-MRS to evaluate GABA in PI, both of which demonstrated reductions in brain GABA.10,11 The initial study10 evaluated GABA over a large brain area that included cortical and subcortical structures. A replication study in a different cohort of PI subjects, using single-voxel 1H-MRS, demonstrated reductions in GABA of 30% and 21% in the occipital and rostral anterior cingulate cortex (rACC), respectively.11 Because of the discrepancies between these studies and the recent work by Morgan and colleagues, careful analysis related to MRS methodologies, differences in study participants, and timing of MRS acquisition may guide further scientific inquiry into the role of GABA in PI. 1H-MRS has very limited spatial resolution compared to other neuroimaging modalities, and therefore, only the aggregate quantity of GABA in a relatively large brain region can be ascertained. Although the sampling region in our preliminary study10 is not comparable to the work by Morgan et al.,1 the more recent replication study similarly evaluated a voxel in the occipital cortex.11 However, closer scrutiny of the voxel placement suggests that these voxels (of different dimensions) had limited overlap, with the voxel by Morgan and colleagues1 being positioned more posteriorly. Also, reductions of GABA in the rACC11 cannot be compared to results from Morgan et al., which focused exclusively on the occipital cortex, likely necessitated by the use of a surface coil for spectral acquisition, which provided excellent signal-to-noise, but could not be utilized to evaluate GABA in deeper brain structures below the cortical surface. Both studies1,11 were performed at experienced spectroscopy centers using a high field magnet (4 Tesla) and J-difference editing techniques. However, there are some subtle methodological differences between the studies that are important to consider. Morgan et al.1 utilized image-selected, in vivo spectroscopy (ISIS) for localization, while our recent replication study11 used point resolved spectroscopy with MEGA suppression (MEGAPRESS), which has evolved into the most commonly used and accurate spectroscopic method for quantifying GABA in vivo.12 Both studies normalized GABA to total creatine (Cr), with neither study demonstrating significant differences in Cr between PI and comparison groups.1,11 However, Morgan and colleagues do not report tissue segmentation data, which could theoretically affect results since GABA is found in greater concentration in gray than white matter.13 At the subject level, both studies evaluated unmedicated PI subjects who reported no prior history of depressive illness.1,11 Morgan and colleagues also considered a first-degree relative with depression exclusionary, which may have affected results given well-established decrements in occipital GABA in major depressive disorder (MDD).14,15 However the heritability of reduced occipital GABA is unclear given recent negative findings in asymptomatic subjects with a positive family history of depression.16 The discrepancies between studies may have also been related to differences in control groups, with controls in our replication study11 composed of more consistently good sleepers. However, because of divergent methodologies to objectively quantify sleep quality (actigraphy versus ambulatory polysomnography), systematic differences among control subjects between studies remains speculative. It is also unclear whether Morgan and colleagues1 accounted for changes in GABA across the menstrual cycle, which may be an important uncontrolled covariate in their design, given prior reports of the GABAergic system being modulated by menstrual phase.17 It seems more likely that the findings of both studies are equally valid, and that other factors may have influenced di vergent findings. Of particular import is that Morgan et al.1 collected MRS in the evening, while our acquisitions occurred in the morning.11 Although a recent study has demonstrated the diurnal stability of occipital GABA in healthy individuals,18 there is also evidence suggesting that the timing of 1H-MRS may be important in other sleep disorders.19 Thus it is possible that occipital GABA levels in PI may be variable across the day. Based on the results of these studies, further evaluation of the effects of time of day on GABA in PI is clearly needed. The findings from all of these recent 1H-MRS studies1,10,11 provide important insights into the neurobiology of insomnia. However, our understanding of the role of GABA in PI will not advance without further research regarding the mechanisms through which regional changes in brain GABA relate to measures of hyperarousal. Research in MDD, which is characterized by hyperactivity of the hypothalamic-pituitary-adrenal (HPA) axis, has consistently demonstrated 1H-MRS-derived reductions in cortical GABA.14,15 However, only one study, using PET, has attempted to link the GABAergic system in MDD with physiologic hyperarousal, demonstrating reduced [11C] flumazenil binding in the temporal lobe, which inversely correlated with HPA axis activity20—again supporting the notion that a deficient GABAergic system is associated with CNS hyperarousal. Further studies that focus specifically on insomnia and combine 1H-MRS with other modalities (EEG spectral analysis, metabolic rate, etc.) may ultimately provide deeper insights into this conceptual framework. In the interim, we applaud Morgan and colleagues for their thought-provoking findings, which are certain to stimulate further research in this area.