Currently, many studies are performed to assess structural or functional brain connectivity. These studies showed an efficient small-world organization of the brain, which is a requisite adequate for cognitive function^[1]. Furthermore, several cognitive and neurological disorders such as Alzheimer’s disease, schizophrenia, or epilepsy show altered brain network characteristics^[1,2]. However, neuronal activity, which underlies brain function, encompasses action potentials in the neurons and synaptic transmission between those neurons through neurotransmitters. None of these biological processes are directly measurable with functional or diffusion MRI, but neurotransmitters are measureable using MR spectroscopy and might therefore give more direct information about neuronal functioning. Previous studies have already shown associations between functional connectivity and glutamate and GABA concentrations^[3], but these studies only assessed local variations in neurotransmitters. In this study, we assess global neurotransmitter variations and propose the concept of ‘neurotransmitter networks’.

Fifteen healthy participants were included. Each participant underwent 7T MRI (Siemens Magnetom), comprising a T1-weighted image (MP2RAGE, TR/TE 4500/2.39ms, TI1/TI2 900/2750ms, FOV 173x230x230mm³, cubic voxel size 0.9mm³) and a MRSI image (sLASER^[4], cFOCI pulses^[5], TR/TE 5520/38 ms, VAPOR water suppression, FOV 150x150x100mm³, voxel size 9.4x9.4x12.5mm³). Metabolite spectra were analyzed using LCModel (version 6.3-1B) with a simulated basis set of 21 metabolites (Figure 1). Spectra were excluded if the SNR was below 20, if the Cramér-Rao lower bounds (CRLBs) of N-acetylaspartate was higher than 3, the CRLB of total Creatine was higher than 10, or if the CRLB of GABA or glutamate was higher than 15. The T1-weighted image was registered to MNI space, in which an atlas was defined to divide the brain into thirty brain areas (the four lobes divided into white and grey matter, left-right, and fourteen subcortical brain regions). The mean neurotransmitter concentrations were measured in these brain regions (relative to creatine, Figure 2), and were corrected for grey and white matter content^[6].

A number of region pairs display correlated neurotransmitter concentrations over the group, whereas this correlation is absent in other region pairs (Figure 3). If an increased neurotransmitter concentration in one region is accompanied with an increased/decreased concentration in another region, we may assume that these regions support concerted brain activity and are therefore considered ‘connected’^[7]. Thus, connectivity was defined as correlated neurotransmitter concentrations between different brain regions among subjects. To obtain the neurotransmitter connectivity matrix, the Pearson’s correlation coefficient between each pair of regions was calculated. The connectivity between pairs of areas was set to 1 if the correlation was significant (p<0.05) and 0 otherwise. The clustering coefficient and the path length of this binarized connectivity matrix were calculated with the brain connectivity toolbox^[8]. A normalized clustering coefficient and path length were obtained by dividing these measures by the mean clustering coefficient and path length of 100 randomized matrices. These normalized measures give information about the small-worldness of the network: the ability to combine a high integration as well as segregation in a network^[8].

The neurotransmitter concentrations could be measured reliable in 26 out of 30 brain areas, and only these areas were included for further analysis. Twenty-three percent of the remaining regions showed a significant correlation for the glutamate concentration, while a significant correlation for the GABA concentration was found in 14% of the regions. Both glutamate as GABA networks showed small-world characteristics, with a clustering coefficient larger and a path length approximately equal to those of random networks (Table 1).In this study, a new methodology is proposed to investigate brain connectivity by assessing large-scale neurotransmitter networks. Currently, this concept is being tested in healthy volunteers and patients with epilepsy by assessing the reproducibility and comparing the neurotransmitter networks with functional and structural networks. Future studies are planned to validate the glutamate and GABA measurements.