Noninvasive Characterization of Functional Pathways in Layer-Specific Microcircuits of the Human Brain Using 7T fMRI

Abstract

Layer-specific cortical microcircuits have been explored through invasive animal studies, yet it is not possible to reliably characterize them functionally and noninvasively in the human brain. However, recent advances in ultra-high-field functional magnetic resonance imaging (fMRI) have made it feasible to reasonably resolve layer-specific fMRI signals with sub-millimeter resolution. Here, we propose an experimental and analytical framework that enables the noninvasive functional characterization of layer-specific cortical microcircuits. Specifically, we illustrate this framework by characterizing layer-specific functional pathways in the corticogeniculate network of the human visual system by obtaining sub-millimeter fMRI at 7T using a task which engages the magnocellular pathway between the lateral geniculate nucleus (LGN) and the primary visual cortex. Our results demonstrate that: (i) center-surround inhibition in magnocellular neurons within LGN is detectable using localized fMRI responses; (ii) feedforward (LGN → layers VI/IV, layer IV → layer VI) and feedback (layer VI → LGN) functional pathways, known to exist from invasive animal studies, can be inferred using dynamic directional connectivity models of fMRI and could potentially explain the mechanism underlying center-surround inhibition as well as gain control by layer VI in the human visual system. Our framework is domain-neutral and could potentially be employed to investigate the layer-specific cortical microcircuits in other systems related to cognition, memory and language.

Keywords: center-surround inhibition; cortical layers; corticogeniculate feedback; dynamic directional connectivity; high-resolution 7T fMRI; layer-specific fMRI; magnocellular lateral geniculate neurons; primary visual cortex.

Figure 1

An illustration of the experimental design. The visual stimuli consisted of six varying sizes of rightward drifting Gabor patches (θ = 0.25°, 0.5°, 0.75°, 1°, 2°, 3°) and a white cross fixation. Each Gabor patch was displayed for 5 s, and the intervals were randomized to 9–12 s. The supplementary video file accompanying this report shows the stimulus as seen by the subjects.

Figure 2

Illustration of surface-based laminar analysis. (a) Four laminar profiles overlaid on the original volume: white matter surface (red contour), layer VI surface (blue contour), layer IV surface (green contour), and pial surface (yellow contour). (b) A zoomed version of (a) illustrating the layers and their relative distances: the distance between the white matter surface to other laminar surfaces was as follows—0.21 mm (from white matter to layer VI, 10% of the thickness), 1.04 mm (from white matter to layer IV, 50% of the thickness), and 2.22 mm (from white matter to pial surface, 100% of the thickness). (c) Significant activation (Z > 2.3 and a FDR-corrected cluster significance threshold of p < 0.05) overlaid on inflated cortical surface. The white line shows the contour for the left primary visual cortex. (d) A flat patch consisting of significantly activated clusters (Z > 2.3, threshold at FDR corrected p < 0.05) on layer VI within left primary visual cortex. (e) A flat patch consisting of significantly activated clusters (Z > 2.3, threshold at FDR corrected p < 0.05) on layer IV within the left primary visual cortex.

Figure 3

LGN definition and spatial analysis of top 20% activated voxels in LGN. (a) LGN mask from Juelich Histological Atlas (thresholded at 60%) overlaid onto MNI brain template. (b) The top 20% of activated voxels obtained with the 1° visual angle stimulus overlaid on the left LGN mask (the white region in (b) corresponds to the LGN mask shown in (a)) for one subject. The relative position of activated voxels is calculated as Dx/Mx and Dy/My where these quantities are depicted in (b). (c) The top panel plots the relative position of left and right magnocellular LGN identified in (b) with respect to the center for 20 subjects (red star) and the associated group average (green cross); the bottom panel shows the histological coronal sections of human LGN, the red layers represent magnocellular LGN and blue parts represent parvocellular LGN (referred and modified from [3]).

Figure 4

An illustration of the data processing pipeline. First, we performed surface-based laminar analysis including laminar surface reconstruction and the registration of functional MRI data to the laminar surfaces. Second, we extracted mean time series from activated vertices (voxels in the volume become vertices on a surface) within each laminar surface and the magnocellular LGN ROIs. Third, blind deconvolution was performed to obtain latent neuronal time series. Fourth, we utilized the dynamic MVAR model to obtain dynamic effective connectivity (one directional connectivity matrix for each time point; the connectivity direction is from row to column). Fifth, separation of condition-specific effective connectivity (18 × 40 EC values for each condition) for each path. An example is shown for the path from layer VI to LGN. Finally, we performed one sample T-test to determine paths whose strengths significantly differed from zero (red * indicates significant at FDR corrected p < 0.05).

Figure 5

The BOLD response in magnocellular LGN for Gabor patch stimuli. (a,b) correspond to left magnocellular LGN, and (c,d) correspond to right magnocellular LGN. (a,c) the responses of each subject for each visual stimuli (0.25° dark blue, 0.5° blue, 0.75° cyan, 1° yellow, 2° red, 3° dark red), x axis is subject number, and y axis is the response (the percentage of change); (b,d) the plot of the mean response over all subjects vs. stimulus degree (red line), 95% confidence interval (red shade).

Figure 6

Center-surround inhibition in different layers of the primary visual cortex. Panel (a) shows the results for the left primary visual cortex and panel (c) shows the results for the right primary visual cortex. In both these panels, the mean BOLD response across all subjects (dash line) is plotted on the y-axis and the stimulus degree is plotted on the x-axis. The 95% confidence interval is shown as the shadow around the dash line. Red represents layer IV, and blue represents layer VI. Panels (b,d) show the center-surround inhibition (the larger of the difference in responses (1°–2°) or (1°–3°)) in each subject for the left and right primary visual cortex, respectively. Here, the red * represents layer IV, and blue * represents layer VI.

Figure 7

Dynamic effective connectivity results. (a) One example of the connectivity matrix at a given time point, the direction is from row to column, e.g., the left corticogeniculate feedback pathway from layer VI to LGN corresponds to the first row/third column, and the left feedforward pathway from layer VI to IV corresponds to the first row/second column. (b) An illustration of the neuronal circuits involving LGN and primary visual cortex: the black round shape represents an inhibitory interneuron, the red triangle is a neuron in layer VI of the primary visual cortex, and the green star is a neuron in layer IV. The blue dotted line represents the negative feedback pathway, and the red dotted lines are the feedforward pathways (LGN → IV, LGN → VI, and IV → VI). (c) Mean/standard deviations of effective connectivity values vs. stimulus degree for corticogeniculate feedback pathway from layer VI to LGN. This pathway is only significantly smaller than zero under 2° and 3° conditions. The red star indicates significance at FDR corrected p < 0.05.

Figure 8

Mean/standard deviations of effective connectivity values vs. stimulus degree for feedforward pathway from LGN to layer IV in the primary visual cortex (a), LGN to layer VI (b), and layer IV to VI (c).