The most important finding of this study was a decrease of Glu concentration with age in the predominantly gray matter motor cortex region and a trend of Gln increase with age in the predominantly white matter corona radiata region. Furthermore, there was a strong positive correlation of Glu with Gln, NAA, and Cr in the motor cortex region. In addition, concentrations of Glu and Gln were higher in motor cortex region compared to the predominantly white matter region of corona radiata.
The finding of reduced Glu with older age in the motor cortex is consistent with histological animal studies of increased deficits of the glutamatergic systems in the aging brain [8
]. Most of Glu is found in neurons [28
], with extracellular concentration of Glu normally remaining at relatively low levels (0.1–1 mM), because of its excitotoxicity. In glial cells, the Glu concentration is even lower [38
] and thus, contributes virtually no signal to Glu in 1
H MRS. Therefore, Glu reduction with age may signify deficits in neuronal metabolism or possible neuronal loss or shrinkage during normal aging. NAA, like Glu, is also localized primarily in neurons and is considered to be a putative neuronal marker [27
]. Our finding of reduced NAA in the motor cortex region with normal aging is consistent with previous MRS studies that demonstrated decreased NAA with age in certain regions of the normal brain [4
]. The present study further showed a strong positive correlation between Glu and NAA concentration, which is consistent with the idea that Glu is localized primarily in neurons and that neuronal integrity is vulnerable in the aging process. Detection of differential variations between Glu and NAA concentrations with MRS could potentially provide a better index of neuronal integrity than measurements of each metabolite alone, especially in cases where glutamate excitotoxicity occurs.
In contrast to Glu reduction in motor cortex, Gln concentration was somewhat increased in older subjects in the predominantly white matter region of corona radiata. Gln is localized primarily in astrocytes, where glutamine synthetase considered to be a marker of glial activity [21
]. Therefore, the finding of increased Gln in this region could be indicative of glial (astrocytes) proliferation in white matter with normal aging. If neuronal loss is accompanied by glial proliferation in healthy aging, the increase of Gln in the white matter should be occurring concomitantly with decrease of Glu and NAA in the same region. However, we failed to detect decrease of NAA and Glu in COR region.
In addition to Gln, other metabolites such as Cr, Cho, and m-Ins were also elevated in the COR region of older subjects, consistent with findings from other laboratories [31
]. It should be noted that 1
H MRS detects the overall metabolite concentration without differentiation between specific components (i.e. glial versus neuronal), and it is not clear to what extent Cr and Cho represent glial compartments. In contrast, no age-related NAA and Glu changes were detected in the COR region, consistent with other 1
H MRS studies that also report stable NAA with aging in white matter [5
]. In this context, it is interesting to note that m-Ins concentration decreased in the MC region and increased in the COR region in older subjects compared to the younger group (m-Ins is used predominantly as glial marker in the 1
H MRS literature [3
]). The increase of m-Ins in white matter is consistent with its role as a glial marker [3
] and concomitant Gln increase in the same region, but we have currently no explanation for the reduction of m-Ins in the motor cortex. The observed differential changes of metabolites with age in two regions are most likely affected by white matter/gray matter composition of those regions.
Although a positive correlation between Glu and Gln in the motor cortex seems counter intuitive to our hypothesis of a negative relationship between the two metabolites, with the former representing neuronal pool and the latter, astrocytes/glia, there are at least two possible explanations for a positive correlation. First, since both Glu and Gln have much higher concentrations in gray matter than white matter, partial gray/white matter volumes of voxels in the motor cortex could dominate the relationship. However, after corrections of the metabolites for gray matter composition of the voxels using linear regressions, the positive correlation between Glu and Gln was still significant. Therefore, gray/white matter partial volume effects cannot entirely explain the positive correlation. Second, there are several synthetic pathways for glutamate and glutamine in brain tissue, and some pathways utilize one metabolite as precursor for synthesis of the other. For example one of the pathways for formation of glutamate is via glutamine, which is imported from glia by phosphate-activated glutaminase. Alternatively, glutamine may be formed from glutamate, originating from α
-ketoglutarate in glia [38
]. Therefore, certain synthetic pathways, which link both metabolites together, could lead to a positive correlation between Glu and Gln, which can only be determined via kinetic studies of specific activities for both metabolites.
Independent of age, levels of Glu, NAA, Gln, and Cr were higher in MC region, compared to COR region. Outside the motor cortex, other MRS studies at 1.5 and 4 T [1
] also reported higher Glu, Gln, and Cr concentrations in gray matter than in white matter. Furthermore, a 1
H MRS study on autopsied brains measured higher Glu, NAA, and Cr in gray matter compared to white matter [30
], consistent with the results of this study. However, tissue-related differences of metabolite concentrations may also be regionally dependent, which limits comparisons to other studies that did not specifically measure metabolite concentrations in motor neuron regions.
In this study, cerebral water was used as reference to determine metabolite concentrations in the motor pathways of human brain. The accuracy of this method was assessed with test–retest measurements, which showed coefficient of variations for Glu between 11% and 15%. In contrast, Gln variations were determined to be two to four times larger, reflecting significant limitation in measuring reliably glutamine concentration even at 4 T, presumably due to much lower signal intensity of Gln compared to Glu [22
]. Although comparison with another study at 4 T [1
] is limited since a different area was studied (thalamus), CV values in this study are similar.
A limitation of the study was that no relaxation measurements of metabolites were obtained for each subject and therefore some regional and age-related changes might be due to relaxation effects (mainly longitudinal relaxation, since the contribution of the transverse relaxation with a very short echo time employed here is negligible). Moreover, different T1
relaxations for resonances of the same metabolite were not considered. Therefore, metabolite quantification in this study is limited to the degree that resonances of the same metabolite exhibit different T1
relaxations. However, a differential effect of age or brain region on T1
relaxations of different functional groups within the metabolite is not expected and therefore observations in this study of age and region related variations of metabolite concentrations should not be confounded by this effect. A second limitation was that we did not correct for partial volumes of gray and white matter when calculating metabolite concentration. Performing this correction is important when interpreting metabolite concentrations of motor cortex, since voxels in those regions contained also substantial amounts of white matter. Another limitation is using cerebral water as a reference without considering potential decrease of water concentration in gray matter and white matter with age, as suggested in Ref. [5
]. Although we did not find significant water content correlation with age in our study and therefore did not take this effect into consideration, sensitivity of our methods to detect an effect may have been limited.