This study aimed to test whether the responsiveness of the GABAergic system to modulation (here achieved through brain stimulation) was associated with variation in short-term motor learning behavior and in learning-related brain activity change. A positive correlation was observed between tDCS-induced GABA decrease in primary motor cortex (M1) and degree of motor learning, such that subjects who demonstrated a greater decrease in M1 GABA following stimulation to M1 also showed faster short-term learning. This finding is in line with the hypothesis that LTP-like plasticity within the neocortex is critically dependent on GABA modulation [10–12
]. A negative correlation was also demonstrated between GABA change and learning-related change in fMRI activity within the left M1, such that the greater the decrease in GABA induced by tDCS, the greater the learning-related reduction in fMRI activity. This observation supports the notion that functional plasticity measured with fMRI reflects GABAergic modulation. The functional and anatomical specificity of this relationship is underlined by the finding that, despite testing for a relationship between GABA modulation in M1 and learning-related fMRI change across the whole motor network, the only region showing a significant correlation was localized within the hand area of the primary motor cortex [13
A number of animal studies have demonstrated that a decrease in tonic GABA is essential for LTP-like plastic changes to be inducible within the motor cortex [10–12
]. In humans, a decrease in GABA induced via an ischemic forearm block [14
] results in a facilitation of LTP-like plasticity, observed as an increase both in the degree of motor learning [1
] and in the response to facilitatory repetitive transcranial magnetic stimulation [15
]. In addition, decreases in GABA concentration have been demonstrated using MRS during motor learning [4
] and in the acute phase of recovery after stroke [16
Distinct relationships were found with measures of baseline GABA concentrations. Subjects with higher baseline levels of M1 GABA had slower reaction times and smaller task-related BOLD signal change in left M1. These findings are consistent with previous reports in the occipital cortex, where BOLD signal change in response to a simple visual stimulus correlated with GABA measurements from the primary visual cortex [17
]. The results presented here suggest that this relationship is generalizable beyond the visual cortex, and, further, the lack of correlations between GABA concentration in our visual cortical control voxel and any fMRI or behavioral measures suggests that these relationships are specific to task-relevant cortical areas.
MRS allows accurate quantification of the concentration of neurochemicals within a defined area of cortex, but it does not give direct information as to the synaptic activity within that area and cannot determine where, within the voxel, the changes in neurochemicals occur. However, the close relationship between perisynaptic GABA concentration and vesicular release [18
], and the finding that change in MRS-assessed GABA via pharmacological and physiological interventions is associated with a facilitation of LTP-like plasticity [1, 4
], suggests that measures of GABA concentration have functional relevance.
This study reports on correlations between GABA and behavioral or functional measures, but it cannot infer causality. Further work should aim to relate GABA decrease during learning with the degree of learning achieved, although this is a technically challenging proposition. In addition, a study designed to test the relationship between tDCS-induced changes in GABA during learning and tDCS-induced changes in behavior during learning would make a stronger case for the causality of the relationship between tDCS-induced GABA change and learning demonstrated here.
It may be that our measure of GABA is a surrogate marker for other changes, such as glutamatergic modification, which in turn may be highly correlated with GABAergic change. Although we find no correlations with our MRS measure of Glx, it should be noted that in our experience MRS measures of Glx appear to have lower sensitivity than GABA measures and in particular are not sensitive to changes in glutamate receptor activity or density, important determinants of neocortical plasticity. In addition, a number of neuromodulators, such as dopamine, serotonin, and acetylcholine, are highly relevant in motor learning, and MRS is insensitive to changes in these chemicals. However, taken together with previous pharmacological studies showing modulation of practice-dependent plasticity with GABA modulation [2, 3
] and the animal literature demonstrating the relevance of GABA to learning [19, 20
], we believe that the evidence presented here suggests that GABA is important in motor learning in healthy adults.
Because we wished to investigate the relationship between the responsiveness of the GABA system to modulation by tDCS and motor learning, we did not include a sham stimulation condition, but rather included a condition using a control voxel. A previous study using identical stimulation and MR parameters showed that there is no modulation of GABA with sham stimulation [5
The finding that the responsiveness of the GABA system to modulation has a strong relationship with motor learning is suggestive of a possible relevance of GABA in an LTP-like synaptic plasticity in human motor learning. In addition, it offers a strong rationale for the use of interventions, such as anodal tDCS, that decrease GABA in a localized region in conjunction with rehabilitation in the context of stroke, in line with recent findings in an animal model [21