9.1 Background

Glutamate (Glu) and gamma-aminobutyric acid (GABA) are the main brain excitatory and inhibitory neurotransmitters. These neurotransmitters are involved in neural migration, differentiation, and synaptic plasticity (1–6). There are accumulating evidence in literature that the neurobiology of mood disorders may arise from an imbalance between excitatory Glu (7–11) and inhibitory GABA (12–15) in key brain regions involved in mood regulation, such as the anterior cingulate cortex (ACC). Glu and GABA dysfunctions have been reported in patients with mood disorders based on several lines of evidence, such as abnormalities in cerebrospinal fluid (CSF) (16–21) and postmortem brain tissue (PMBT) (1–6). Additionally, glial cells have been reported to be reduced in many brain (including prefrontal cortex (PFC) and anterior cingulate cortex) areas of subjects with bipolar disorder (BD) according to PMBT studies (9, 22). In major depressive disorder (MDD), Glu levels have been reported to be elevated in plasma (12–15) and PMBT studies (16–21).Thus, evidence has encouraged the development of modern neuroimaging techniques that allow the in vivo measurement of Glu and GABA.

9.2 Glutamate

The Glu hypothesis of depression was proposed in the 1990s, when antagonists of the N -methyl-D-aspartate (NMDA) receptor, an ionotropic glutamate receptor, were found to possess antidepressant-like mechanisms of action in mice (23). More recently, it has been reported and replicated that the infusion of low-dose ketamine, an NMDA receptor antagonist, is associated with robust decreases of depressive symptoms, mainly suicidal thoughts, in depressed patients (24–26). This way, the balance between brain levels of Glu and GABA is hypothesized to be as crucial for achieving and sustaining euthymia in the treatment of mood disorders. However, only recently the development of neuroimaging techniques allowed the precise measurement of these metabolites.

Glu is the most abundant excitatory metabolite in the brain (27), and it is continuously recycled between neurons and glial cells. Excess Glu causes excitotoxicity and apoptosis (28), which is also associated with higher levels of intracellular calcium (Ca⁺²) and production of mitochondrial reactive species. (29, 30). After neuronal Glu is released into the synapse, it is taken up by astrocytes and converted to glutamine (Gln) by Gln synthase (GS) (31) via ketoglutarate from the tricarboxylic acid (TCA) cycle. Moreover, astrocytes deliver energy sources, including glucose and lactate, based on neuronal energy requirements; thus, there is a close link between the glutamatergic system and brain energy metabolism via astrocytic functions (32). The cycle involving neuronal Glu release from neurons and its resynthesis from Gln inside astrocytes is known as the Glu–Gln cycle (33). Glu does not cross the blood–brain barrier, and its concentration in the cellular and extracellular fluid is maintained at lower levels by the Glu–Gln recycling across neurons and astrocytes (34). Since absolute measures of Glu and Gln do not reflect the constant flux through the Glu–Gln cycle, it has been suggested that the Glu/Gln ratio might be more sensitive to measure changes in the Glu–Gln cycle rather than either metabolite alone (35, 36). Thus, the Glu/Gln ratio may potentially provide insights into glutamatergic activity, and changes in this ratio could be interpreted as a measure of Glu–Gln cycle rate. Some groups have theorized that the Glu/Gln ratio might reflect the flux through the Glu–Gln cycle and can serve as an overall indicator of glutamatergic transmission activity (35–42).

9.3 GABA

GABA is the most abundant inhibitory neurotransmitter in the central nervous system (CNS), and it is synthesized in GABAergic neurons from Glu by the enzyme glutamic acid decarboxylase (GAD) (43, 44). GABA synthesis in the human brain depends greatly on the GAD1 gene, whose expression and protein levels have been reported to be reduced in different brain regions of BD and schizophrenia patients, including in the ACC (5, 45–48). After its release for neurotransmission, GABA is taken up by both GABAergic neurons and by astrocytes. In neurons, it may be stored in vesicles and reused for neurotransmission or it is degraded by the mitochondrial enzyme GABA transaminase (GABA-T) and enters the TCA cycle, being recycled to glutamate and then GABA again, a process known as the GABA shunt. In astrocytes, the GABA is also converted to glutamate and subsequently to glutamine. Glutamine may enter into the TCA cycle or it may be released to neurons for glutamate synthesis as explained earlier (49). Therefore, in the normal brain, there is a physiological Glu–Gln-GABA cycle associated with neurotransmission and TCA (50).

9.4 Magnetic Resonance Spectroscopy

Proton magnetic resonance spectroscopy (¹H-MRS) is a noninvasive technique particularly useful to assess the brain neurometabolic profile by measuring the content of several metabolites including glutamate and GABA (52) (Figure 9.1). Briefly,¹H-MRS relies on the same technique used for magnetic resonance imaging (MRI). MRI signal is based on the magnetic properties of the hydrogen nuclei. In MRI, most of the signal arises from water, which is the most abundant molecule, and the effect of chemical shift can be neglected in most cases. In ¹H-MRS, however, this effect is explored in order to distinguish between the different molecules or metabolites. For this purpose, the strong water signal in the brain voxel is first saturated, and the residual signal is then displayed as a spectrum, that is, as a function of frequency, in order to identify all signal frequencies present, which will be then assigned to different metabolites ¹H. In ¹H-MRS frequency, values are given in parts per million (ppm), a unit which is independent of the magnetic field strength of the scanner and makes its interpretation easier.

Figure 9.1 One-dimensional projection of the acquired JPRESS spectrum. Measured spectrum in blue, fitted in red, and residual in green.

¹H-MRS technique is able to detect nearly 80% of brain Glu, with its most prominent peak at 2.34 ppm in the spectrum that arises from the methylene protons near the carboxy-terminal of Glu (52). Gln concentration is estimated to correspond to 40–60% of Glu, but its separate quantification can be obtained only using special ¹H-MRS techniques and/or higher magnetic fields. Glu and Gln are very similar molecules, and for this reason, their ¹H-MRS signal, represented by broad and complex peaks, overlaps very strongly. A detailed review of the techniques available to detect Glu and Gln can be found in (49). For this reason, most studies refer to the observed signal as ‘‘Glx,” which corresponds to a greater proportion of Glu, but also Gln, GABA, and glutathione (GSH). Since Glu and Gln are the major contributions for “Glx,”it has been considered a proxy of total glutamatergic neurotransmission (8, 49).

In ¹H-MRS spectra, GABA is commonly observed at about 2.28 ppm and is partially overlapped by the Glu peak, being considered nearly 15–20% of Glu concentration (27). Due to such overlapping signals, GABA cannot be quantified with conventional brain ¹H-MRS acquisitions, but it is possible to measure it by using magnetic fields with strengths higher than 3.0 T and special pulse sequences such as the J-resolved and J-difference editing sequences (49).

The ACC has been considered a center of integration of cognitive and affective neuronal connections (53). It has been the most studied brain region in affective disorders in ¹H-MRS studies. Considering the key role of ACC for affective and cognitive regulations as well as its connective functions between frontolimbic structures (54), there are plenty of studies showing structural and functional abnormalities in the ACC in BD. Reduced ACC gray matter volume has been demonstrated since the work performed by Drevets et al. (55) and then extensively confirmed by several authors (54, 56–58). More recently, Hibar et al. (59) have corroborated this finding in the largest BD sample assessed to date.

9.5 Major Depressive Disorder and Glutamate

Since the first (23) to the most recent studies (24–26), a consistent body of evidence has linked depression physiopathology to Glu as confirmed by a recent meta-analysis (60)(Table 9.1). Most studies have reported a decreased level in MDD patients or no differences in relation to healthy controls (HCs). Recent ¹H-MRS meta-analyses that have addressed glutamatergic alterations in MDD across several brain regions have also reported similar results (61, 62). Godfrey et al. (61) investigated Glx data in multiple brain voxels in 520 MDD patients compared to 501 controls across 24 studies, as well as Gln in 444 MDD patients and 420 HCs and found no differences in either Glx or Gln between MDD and HC. However, when they restricted their analysis to the ACC of 232 MDD patients compared to 226 HCs across 12 studies, Glx was found significantly lower in MDD, although no difference was found for Glu, which remained similar to HC (Table 9.1). Such discrepancies in Glx concentration in whole-brain and specific voxels such as the ACC might be explained by the fact that the former assessment may detect changes in areas not directly related to the etiology of depression. In contrast, changes in neurometabolites in the areas that regulate the cognitive and emotional behaviors such as the ACC (53) are more likely to be implicated in the pathophysiology of MDD. Therefore, more regional and hypothesis-driven ¹H-MRS studies assessing Glu are recommended for MDD.

Additionally, Moriguchi et al. (62) performed a meta-analysis of Glx alterations in the medial PFC (mPFC) of 502 MDD as compared to 408 HC subjects and found significantly lower Glx levels, while no changes were observed for Glu or Gln. A subgroup analysis revealed that Glx was decreased in the mPFC of medicated patients with antidepressants compared to controls, and no difference was observed between unmedicated patients and controls (62). Previous meta-analyses have reported decreased Glx, but not Glu, in the PFC of MDD patients as compared to controls (63), and a decrease of both Glx and Glu in the ACC of MDD patients as well as in the whole brain of those under current depressive episodes (64). Overall, there is converging evidence of decreased Glx levels in frontal areas in MDD as compared to controls.

9.6 Bipolar Disorder and Glutamate

¹H-MRS studies have documented increased Glx in several brain regions in BD such as: ACC (65–70), dorsolateral PFC (71), basal ganglia (65), hippocampus (40, 65, 66, 72), and occipital cortex (65, 72). Indeed, previous reviews (8, 73) and meta-analyses (74, 75) have confirmed that elevated Glx is likely the most striking neurometabolic abnormality in BD subjects, particularly in frontal areas in all mood states (74, 75). Among these brain regions, the ACC is an important area linked to mood regulation (76, 77) and shows widespread functional and structural abnormalities in patients with BD (78)(Table 9.2). Most studies have found elevated Glx or Glu levels in BD or no differences in relation to HC. Increased ACC Glu levels seem to be a trait marker intrinsic to BD (79, 80) since several authors have reported this finding in early-onset (81) and medication-free patients (65, 66, 81). Besides, elevated Glu and/or Glx levels have also been reported across different mood states in the ACC but particularly in the depressed (65, 66, 81) and euthymic (67–70) mood states. Indeed, ¹H-MRS studies in BD type I euthymic patients indicate Glu cycle metabolite abnormalities and also suggest increased Glx and Glu (67, 68, 72, 82, 83) as well as increased Gln (36, 68, 69) within at least three different brain regions processed with different ¹H-MRS sequences (Table 9.2).

Glutamatergic abnormalities in BD also seem to stem from an uncoupled neuron–astrocyte relationship as changes in the Glu–Gln cycle have been documented. Since Glx represents the sum of Glu and Gln signals, studies reporting increased Glx could reflect increased Glu, increased Gln, or even elevated levels of both of these metabolites. Specific ¹H-MRS studies reporting Gln measures are scarce in BD. While Ongür et al. (40) reported higher Gln/Glu ratio in the ACC of manic patients, considering it a measure of glutamatergic activity, Moore et al. (84) found that BD children medicated with anticonvulsants, antipsychotics, and antidepressant had higher levels of Gln than medication-naive patients. Similarly, Soeiro-de Souza et al. (36) and Kubo et al. (69) revealed increased levels of Gln in adult euthymic medicated samples treated with combinations of lithium, anticonvulsants, antipsychotics, and antidepressant but the former study revealed higher Gln levels in their subsample under anticonvulsant treatment in relation to the other group class (Table 9.2). Additionally, Frye et al. (66) observed an increase of Gln in response to lamotrigine treatment and Brennan et al. (35) reported an increased Gln/Glu ratio in BD depression after treatment with riluzole, an antagonist of glutamatergic receptors. Although several studies showed that the elevated Glu concentration remained higher in BD patients in relation to controls even when the influence of medication was statistically removed from their analyses (67, 69, 70), anticonvulsants may likely modulate the conversion of Glu into Gln, whatever may underlie their mood stabilizing mechanism of action.

In fact, mood stabilizers have been shown to modulate the Glu concentration in the brain through multiple mechanisms involving the regulation of synaptic Glu uptake, receptor activity, and intracellular signaling cascade functions (85). While anticonvulsants have been shown to decrease the glutamatergic neurotransmission (85), lithium has been reported to both increase (82, 86, 87) and decrease Glx levels (85, 88), whereby this latter finding was corroborated by a recent systematic review(89).

9.7 Major Depressive Disorder and GABA

GABAergic deficit hypothesis (90) has been proposed for MDD based on several lines of evidence indicating that GABA is reduced in plasma (91, 92), CSF (93), and brain tissue (94) of MDD subjects as compared to HC. Similarly, recent meta-analyses on GABA levels in MDD have consistently shown low brain GABA levels relative to HC in several brain regions such as occipital cortex, parieto-occipital (PO) cortex, ventromedial PFC, and ACC (95–97). Importantly, the separate comparison between depressed and remitted MDD subjects showed that the former revealed significantly lower GABA levels than the latter, which achieved similar levels to controls (96). Similarly, Godfrey et al. (61) have assessed in their meta-analysis GABA levels in 356 MDD patients as compared to 366 HC and found significantly lower GABA levels in MDD patients. When the authors restricted their analyses to the ACC (118 MDD patients and 97 HCs), GABA levels remained lower in MDD patients. However, remitted MDD patients showed similar GABA levels to HC. Accordingly, Table 9.3 demonstrates studies that addressed GABA levels in the ACC of MDD in relation to HC.

Changes in GABA levels are implicated not only in the etiology of MDD (90, 98) but also in its recovery; thus, it is considered a state-dependent rather than a trait marker of MDD (61, 95, 97). In fact, normalization of GABA levels in MDD has been reported in response to treatments(61) with selective serotonin reuptake inhibitors (99) and ECT (100). In this way, GABA appears as a promising tool to assess the treatment response in MDD. However, the correlation between lower GABA levels and depressive symptomatology remains poorly understood. It has been demonstrated in animal models that chronic stress causes a presynaptic GABA decrease and downregulation of postsynaptic GABA-A receptors (101). Such a GABA decrease in frontal areas might result in a dysregulation of excitatory and inhibitory neurotransmission, leading to alterations in mood and cognition. However, depressive episodes are per se stressful events for those who suffer from MDD and the lower GABA levels might also be a result rather than a cause of the disease. Longitudinal studies assessing GABA levels along several mood episodes would better clarify this issue.

9.8 Bipolar Disorder and GABA

Previous reviews and meta-analyses did not find significant differences in GABA concentration in BD in relation to HC (95, 96). Chiapponi et al. (95) investigated five studies (Table 9.4) about GABA levels in the brain and BD and concluded that given the heterogeneity of the studies it was not possible to state that GABA is different from controls in BD brain. Among these studies, Wang et al. (102) reported higher GABA/Cre ratios in the occipital and PFC regions in a sample of sixteen medicated BD subjects (euthymia, mania, and depression)(102) while Bhagwagar et al. (72) examined the ACC of sixteen unmedicated “recovered” BD patients and eighteen controls and found lower GABA/Cre ratios in the BD group (72). In contrast, Kaufman and colleagues (103) examined the PO and thalamic regions of thirteen BD patients (euthymia, mania, and depression) and eleven controls, reporting no between-group differences (103).

Similarly, among the fewer studies that have examined GABA levels in the ACC of BD patients in relation to HC (Table 9.4), it has been reported increased levels in BD or no significant differences. While Brady and colleagues (13) examined fourteen euthymic BD type I patients and fourteen controls and found higher GABA/Cre ratios in the ACC and also PO cortex of BD patients compared to controls (13), the other studies that measured GABA have reported no differences in these metabolite levels compared to HC in both the basal ganglia (104) and ACC (36, 105). Such contradictory results may be explained by variations in ¹H-MRS acquisition methods, brain regions investigated, metabolite quantification and normalization strategies, sample characteristics, and medication status, hampering interpretation of these conflicting findings.

Additionally, there is neither any follow-up study assessing GABA across mood states (hypomania, mania, depression, and euthymia) in BD nor comparing GABA levels between bipolar and unipolar depression (97), which would provide clues on the role of this metabolite in these disorders. However, there are some potential explanations for contrasting between GABA data for MDD and BD. First, there is a larger number of studies that assessed GABA in MDD as compared to BD. Second, it is still a challenge to derive a reliable GABA signal from ¹H-MRS spectra and disentangle it from the macromolecular signal, which requires a special ¹H-MRS acquisition technique (106). Third, most patients in BD studies were medicated with a combination of lithium, anticonvulsants, antidepressants, and antipsychotics. Since anticonvulsants, benzodiazepines, and antidepressants are known to modulate GABA levels (99, 107), the conflicting findings may likely be influenced by the medication effect. In contrast, as observed for the ACC (Table 9.3), most studies that assessed GABA in MDD, the patients were medication-free, thus providing less biased results.

9.9 Concluding Remarks and Clinical Relevance

There is a considerable body of evidence of glutamatergic and GABAergic abnormalities in mood disorders, likely due to an excitatory/inhibitory imbalance. The Glu and GABA balance in the developing brain is implicated in key processes such as neural migration, differentiation, and synaptic plasticity (2, 4), and consequently the normal adult brain function (108). Thus, it has been proposed that the mood instability and cognitive impairments observed in mood disorders, namely MDD and BD, may result from an excitatory/inhibitory imbalance in cortical regions involved in affect and cognitive regulation as the ACC (109, 110).

Glutamatergic abnormalities have been consistently reported for both BD and MDD. Increased Glu levels have been reported in several cortical regions in BD (8, 73–75), which could be considered a trait marker for this disorder since it has been observed in all mood states. The most accepted theory for the deleterious effects of Glu hyperactivity to cells is based on evidence that a supraphysiological activation of the glutamatergic receptors may result in an increased intracellular Ca²⁺ concentration, leading to the activation of Ca²⁺-dependent enzymes, causing mitochondrial Ca²⁺ overload, oxidative stress and stimulation of apoptotic pathways, resulting, ultimately, in neurotoxicity and neuronal death (74, 111). Therefore, cortical reductions in frontal regions observed in BD (59) may result in part from such hyperglutamatergic state, leading to cognitive and affective dysfunction. Conversely, lower Glu levels have been documented in frontal regions in MDD such as the ACC (61) and mPFC (61, 62), which is considered an indicator of severity and poorer outcome (110). Thus, it has been suggested that successive depressive episodes may result in lower synaptic strength and consequent reduced Glu neurotransmission.

GABAergic abnormalities have also been implicated in mood disorders, particularly in MDD. Reduced GABA levels have been found in several cortical regions in MDD, and its recovery has been associated with the treatment response, thus, it is considered a state-dependent trait marker of MDD (95–97). Such reduced GABA levels observed in MDD has been associated with chronic stress (101) and maybe either a causative factor or a consequence of consecutive depressive episodes. On the other hand, GABA neuroimaging data have provided less positive consistent results for BD than MDD, likely because the positive studies in MDD were executed with medication-free patients, which is harder to achieve in BD than in MDD.

Overall, there is converging evidence of altered Glu levels in both BD and MDD and GABAergic abnormalities only in MDD. Thus, changes in the excitatory/inhibitory neurotransmission, particularly in frontal regions as the ACC, seem to be associated with mood and cognition alterations observed in both disorders. However, the methodological variability across studies (different voxel sizes and locations, acquisition, and post-processing techniques in multiple mood states and BD subtypes in the same sample and possible medication effects) precludes any further conclusions from being drawn. Besides, very few studies have examined simultaneously and longitudinally both neurometabolites Glu and GABA in mood disorders, which would enable us a better understanding of their roles in the neurobiology of BD and MDD and the clinical implications.

Since the frontal areas such as the ACC are associated with cognitive and affective function, and it has been demonstrated reductions of these cortical regions in BD (59), we may infer that the hyperglutamatergic state may underlie the structural and functional changes observed in this region. Additionally, more research is required to understand the longitudinal dynamic changes in Glu levels across the different mood states, would enable us a better understanding of the connections of glutamatergic alterations and BD symptoms. On the other hand, it seems that GABA neuroimage has less positive data in BD than MDD, maybe because the positive studies in MDD were executed with medication-free patients, which is harder to achieve in BD than in MDD.

Further ¹H-MRS studies at higher magnetic fields with more sensitive sequences should investigate specific BD subtypes and mood states to increase understanding of the Glu system in BD and help develop novel pharmacological approaches based on the glutamatergic system. Considering that previous studies in patients during mania (40, 112) as well as unipolar and bipolar depressions(65, 66, 110) have indicated altered Glu system metabolites in mood disorders, we may hypothesize that such abnormalities might be a putative neurobiological endophenotype for mood disorders. However, ¹H-MRS studies are needed to confirm this hypothesis by comparing not only affected and unaffected subjects but also unaffected first-degree relatives, which so far has not been the case (113).