Hypothesis and experimental design. (a) Schematic of hypothesis. The inhibition mechanism centered on MT+ GABA, including the molecular level: the GABAergic inhibition in MT+ (green circle), brain connectivity level: hMT+ - frontal functional connectivity (blue circle), and behavior level: hMT+ specific surround suppression of visual motion ( red circle), contributes to the visuo-spatial component of gF (3D domain, yellow circle). (b) Schematic of experimental design. Session 1 (rectangle box of short line) was the functional MRI and MRS scanning at resting state. Session 2 (rectangle box of solid line) was another region of MRS acquisition. In the two sessions, the order of MRS scanning regions (hMT+ and EVC (primarily in V1)) was counter balanced across participants. There was a structure MRI scanning before each MRS data acquisition. The interval between the two sessions was used for behavioral measurement (rectangle box of dotted line): block design task (BDT) and psychophysical task - motion discrimination. Sold lines indicate the experiment sequence.

hMT+ localizer scans and hMT+ MRS VOI placement. (a) Single task block designs. First: a cross fixation on the center of the screen (10s). Second: a moving grating (2°) toward left last 10s. Third: the grating keeps static for 10s. Fourth: the grating moves toward right last 10s. The localizer scans consist of 8 blocks. (b) hMT+ location and MRS VOI placement. The upper template is the horizontal view. The lower templates from left to right are coronal and sagittal views. The warm color indicates the overlap of fMRI activation of hMT+ across 14 subjects, the cold color bar indicates the overlap of MRS VOIs across all subjects.

The left hemisphere EVC (primarily in V1) MRS scanning ROI. The color bar indicates the overlap of the MRS VOIs across all subjects.

MRS spectra and the relationships between GABA levels and SI / BDT. (a) Example spectrum from the hMT+ voxel of one participant. The first line is the LCModel fitting result of all metabolites, and the following lines show the Glu and GABA spectra fitting with LCModel, and then the baseline. (b) Individual participants fitted GABA MRS spectra from the hMT+ (top) and EVC (primarily in V1) (bottom) voxels from baseline measurement. The colors of the GABA spectra represent the individual differences of BDT. The color bar represents the scores of BDT. (c) and (d) Pearson’s correlations showing significant positive correlations between hMT+ GABA and BDT scores (c), between hMT+ GABA and SI (d). (e) Peason’s correlation showing significant positive correlations between SI and BDT. (f) Peason’s correlation showing significant negative correlations between BDT and duration threshold of small grating. (g) No correlation between BDT and duration threshold of large grating. The ribbon between dotted lines represents the 95% confidence interval, and the black regression line represents the Pearson’s correlation coefficient (r). GABA and Glu concentrations (Conc.) are absolute, with units of mmol per kg wet weight (Methods).

Relationships between GABA concentration in EVC (primarily in V1) and BDT/SI. (a) There is no significant relationship between BDT and GABA concentrations in EVC (primarily in V1) region. (b) There is also no obvious correlation between SI and GABA concentrations in EVC (primarily in V1) region. GABA concentration (Conc.) is absolute, with units of mmol per kg wet weight (Methods).

Individual Glu MRS spectra from hMT+ / EVC (primarily in V1) regions and relationships between BDT and Glu concentrations in hMT+ / EVC (primarily in V1) regions.

(a) Individual participant fitted Glutamate MRS spectra from the hMT+ (top, n = 31) and EVC (primarily in V1) (bottom, n = 28) voxels from baseline measurement. The colors of the Glutamate spectra represent the individual differences of BDT. The color bar represents the scores of BDT. (b) There is no significant relationship between BDT and Glu concentrations in hMT+ region. c There is also no significant correlation between BDT and Glu concentrations in EVC (primarily in V1) region. Glutamate concentration (Conc.) is absolute, with units of mmol per kg wet weight (Methods).

Correlations between GABA and Glu concentrations in hMT+ and EVC (primarily in V1) regions. (a) There is significant correlation between GABA and Glu concentrations in hMT+ region. (b) The levels of GABA also significantly correlate with Glu in EVC (primarily in V1) region. GABA and Glu concentrations (Conc.) are absolute, with units of mmol per kg wet weight (Methods).

Two linear correlations. (a) Significant positive correlation between suppression index and BDT scores (took out two outlines). (b) BDT also significantly correlates with GABA concentration in hMT+ region (without two outliers, having the similar result shown in Figure 3d).

FC of voxels showing significant correlation with BDT scores across subjects in frontal cortex.

FC of voxels showing significant correlation with BDT scores across subjects in whole brain.

FCs of voxels showing significant correlation with SI across subjects in frontal cortex.

FCs of voxels showing significant correlation with SI across subjects in whole Brain.

Significant FCs from connectivity-behavior analyses in a priori search space. The seed region is the left hMT+. The significant FCs are obtained from a priori space (frontal cortex). (a) The significant FCs obtained from connectivity-BDT analysis. Single voxel threshold P < 0.005, adjacent size ≥ 23 (AlphaSim correcting, Methods). (b) The significant FCs obtained from connectivity-SI analysis. Single voxel threshold P < 0.005, adjacent size ≥ 22 (AlphaSim correcting, Methods). Positive correlations are shown in warm colors, while negative correlations are shown in cold colors. The paired deep green, deep pink, light green circles on (a) and (b) indicate the overlap regions in left BA6, right BA6, right BA46(DLPFC) between Connectivity - BDT analysis and Connectivity – SI analysis, respectively.

Significant FCs searched from connectivity-behavior (BDT/SI) analyses in the whole brain. The seed region is the left hMT+. The significant FCs are obtained from the entire brain search, single voxel threshold P < 0.01, adjacent size ≥ 37 voxels (AlphaSim correcting, Methods). Positive correlations are shown in warm colors, while, negative correlations are shown in cold colors. (a) the significant FCs obtained from connectivity-BDT analysis. (b) the significant FCs obtained from connectivity-SI analysis.

Significant FCs searched from connectivity-behavior (BDT) analyses in the frontal (a) and whole brain (b). The seed region is the left V1. (a) The significant FCs are obtained from the frontal search, single voxel threshold P < 0.005, adjacent size ≥ 23 (AlphaSim correcting, same methods for hMT+). (b) The significant FCs are obtained from the entire brain search, single voxel threshold P < 0.01, adjacent size ≥ 37 voxels (AlphaSim correcting, same methods for hMT+). Only positive correlations were detected, shown in warm colors.

Correlations between FC in Table 1 and GABA/Glu concentrations in hMT+.

Correlations between FC in Table supplement 1 and GABA/Glu concentrations in hMT+

Correlations between FCs in Table supplement 2 and GABA/Glu concentrations in hMT+.

Correlations between FCs in Table supplement 3 and GABA/Glu concentrations in hMT+.

Local hMT+ GABA acts on SI and BDT via global hMT-frontal connectivity. (a) Significant negative correlation between the FC of left hMT+ - right DLPFC (BA46) and hMT+ GABA (FDR correction). (b) Significant positive correlation between the FC of left hMT+ - right (pre) motor cortex (BA4/6) and hMT+ GABA (FDR correction). In (a) and (b), left: the significant FCs obtained from connectivity-BDT analysis; right: the significant FCs obtained from connectivity-SI analysis. (c) Significant pathways: hMT+ GABA→ FC (left hMT+ - right BA46, negative correlation) → SI (negative correlation) → BDT (positive correlation). This pathway can explain 34% of the variance in BDT. (d) Significant pathways: hMT+ GABA → FC (left hMT+ - right BA6, positive correlation) → BDT (positive correlation).The bolded lines represent the hypothesized mediation effect. The dotted lines represent alternative pathways. *: P < 0.05; **: P < 0.01; ***: P < 0.001.

The alternative serial mediation models from local hMT+ GABA to global performance of BDT. The pathway that hMT+ GABA was predicted to be associated with SI, followed by the FC of hMT+ - BA 46, and then BDT, did not yield the chained mediation effects on BDT. *: P < 0.05; **: P < 0.01; ***: P < 0.001.

Correlation matrix shows correlations between all measured in our biochemistry and behaviour level. *: P < 0.05, **: P < 0.01, ***: P < 0.001.

Sketch depicting the multi-level inhibitory mechanisms centered on hMT+ GABA contributing to visuo-spatial intelligence. Inhibitory GABA in hMT+ (a sensory cortex, shown in green circle), coupling with the functional connectivity between hMT+ and BA46 (cognitive control core, shown in purple circle), and mediated by motion surround suppression (shown in red circle), contributes to visuo-spatial intelligence (BDT, 3D domain, shown in red and white building blocks). In this sketch, the two-colored parallel arrows show the negative FC between hMT+ and BA46, the colored arrows below the green circle display the inhibition mechanisms centered on hMT+ GABA (2), filtered the irrelevant information in (1) and focused on the efficient visuo-spatial information (3). Black long arrows display the direction of information flow: from input information (1) to visuo-spatial intelligence (4).