Here we report novel differences in 1H-MRS defined levels of metabolites in the ACC and insula measured in the interictal period of migraine patients. Although conventional descriptive statistics yielded no differences on the analysis of the spectra, a LDA demonstrated significant differences between migraine subjects and age-gender matched controls. This type of analysis allows for the determination of discrimination of two or more groups (e.g., migraine vs. healthy) based on Cr-normalized levels of specific metabolites. This analysis separated out a relationship between NAAG and Gln within the ACC and insula during their interictal period. NAAG is the most abundant peptide neurotransmitter in the mammalian CNS  being synthesized exclusively in neurons from NAA and Glu by NAAG synthetase. In addition to its role as a neurotransmitter, NAAG is a source of Glu  and like NAA is thought to play a role as a major osmolyte in the vertebrate brain [65, 66]. Glutamine on the other hand is synthesized exclusively in glial cells from Glu and ammonia by the enzyme glutamine synthetase. Subsequently, Gln is released back into the extracellular space, shuttled back into neurons and converted to Glu by glutaminase. The Glu that is regenerated may then go on to play a direct role in excitatory neurotransmission, packed and stored in vesicles or incorporated into NAAG. An intriguing observation in the present study is the LDA-detected classification of migraine patients and control subjects for two different brain regions based on NAAG and Gln, which are closely linked by this excitatory neurotransmitter system. Interestingly, the ACC and insula LDA plots show oppositely signed gradients, an observation that might be explained by (i) the significant tissue type differences within ACC and insula voxels detected by image segmentation and (ii) the known uneven distribution of Gln and NAAG throughout the brain and within brain tissue type . The measured changes in these excitatory amino acid neurotransmitters (NAAG) and related species (Gln) provide some insights into altered central nervous system (CNS) mechanisms in migraine and may contribute to abnormal CNS processing including changes during the migraine state (e.g., process of central sensitization [68, 69], progressing from acute episodic to chronic/daily migraine  or abnormalities during the interictal period [71–74]. We did not detect direct differences in Glu levels between controls and migraine patients although preferential storage of excess synaptic Glu in the form of Gln and/or NAAG might explain comparable Glu levels within the two cohorts. Note that a previous 1H-MRS study showed decreased cerebellar Glu levels in migraine patients compared to healthy controls  yet similar cortical 1H-MRS findings have not been reported to date.
A growing body of preclinical and clinical data supports the notion of aminergic dysfunction in migraine headache including alterations in both the glutamatergic and glutaminergic systems [31, 32]. For example, NMDA receptor antagonists inhibit cortical spreading depression in the rat brain . Cerebrospinal fluid  and plasma  Glu and Gln levels are increased in chronic migraine patients, although no such data is available for episodic migraine (i.e. our population). It has been postulated that increased brain Glu leads to cortical hyperexcitability typical of migraine [77, 78] and potential pharmacological targets for migraine therapy include the ionotropic (NMDA, AMPA and kainate) and metabotropic glutamate receptor antagonists . The use of a tridimensional personality questionnaire in migraine and tension-type headache clinical sub-populations has shown that glutaminergic dysfunction might also be a specific feature associated with migraine headache . The development of novel pharmaceutics that can modulate the glutaminergic system and block central and peripheral sensitization might be efficacious for treating migraine. It is also worth noting that, although little is known in the literature for a potential role of NAAG in migraine, there may be a potential role for NAAG antagonists (via mGluR3 receptor blockade) for the therapy of migraine.
A number of reports indicate that modulation of the glutamatergic system in the ACC takes place following pharmacological or sensory manipulation. Alterations in ACC neurons may be dependent on prior events that change or modulate neuronal activity. For example, drugs may decrease levels of glutamate in the ACC  and excitatory synapses into the ACC are in part NMDA mediated changes in this region . In addition, amputation of a hind paw digit in rats results in a loss of activity-dependent long-term depression in the ACC  and potentiation of sensory responses . NMDA receptors in the ACC mediates pain-related aversion . Thus, in migraine patients either as a result of intermittent pain or medications, ACC glutamatergic impairment would account for an increase in activation in this region. In data from another report we observe increased sensitivity in the descending modulatory systems in the brainstem in interictal migraine patients vs. controls . In functional imaging studies of pain, activation in the insula is observed and it has been suggested that the region has important contributions to both pain and emotional processing [87, 88]. However, 1H-MRS detected changes in this region in the interictal period have not been reported.
For the present study, we chose to use a 3D localized variant of J-resolved 1H-MRS, a method that has been shown to enhance spectral resolution at several field strengths including 1.5 T [42, 43], 3.0 T  and 4.0 T . Increased spectral resolution is achieved as J-coupled metabolite resonances are effectively spread over a 2D surface whereas uncoupled peaks remain along F1 = 0 Hz. Glutamine contains a single methine (CH; 3.75 ppm) and two methylene (CH2; 2.1 and 2.4 ppm) groups and each proton resonance is split owing to J-coupling effects . It follows that, for 2D J-resolved 1H-MRS data, glutamine shows multiple proton resonances across the 2D surface. In combination with LCModel fitting and GAMMA-simulated basis sets, we use information from the whole 2D datasets and this approach further improves multiple-metabolite quantification of 2D 1H-MRS data. Recently we applied these methods in vivo and demonstrated their utility for reliably measuring brain glutamate and glutamine levels . NAAG, a dipeptide composed of NAA and Glu joined by a peptide bond, also benefits from the 2D 1H-MRS approach. The major resonance of NAAG is its CH3 resonance at 2.04 ppm that appears as a shoulder on the dominating NAA CH3 2.0 ppm peak. In conventional 1H-MR spectra, this chemical shift region is further complicated by underlying J-coupled resonances of Gln, Glu and GABA, and a major advantage of 2D J-resolved data is the fact these J- coupled metabolite resonances are shifted away from the F1 = 0 Hz axis. This yields a cleaner chemical shift region that is essentially comprised of NAA and NAAG CH3 singlet peaks, both of which are more reliably fitted by the described LCmodel template and fitting procedures.
It is certainly true that MRS studies are limited by the relatively low SNR of the spectra and some studies of chronic pain patients have noted larger between group differences. However, the methods that were employed in the present study were designed to allow the detection and quantitation of a larger number of lower concentration metabolites. This has to make observations that would not have been possible using more standard methods. In addition the interictal migraine group may differ from the chronic pain state in that it produces prolonged and continuous brain changes that manifest in profound structural  and functional changes [92, 93].