Translation of monosynaptic circuits underlying amygdala fMRI neurofeedback training

Abstract

fMRI neurofeedback using autobiographical memory recall to upregulate the amygdala is associated with resting-state functional connectivity (rsFC) changes between the amygdala and the salience and default mode networks (SN and DMN, respectively). We hypothesize the existence of anatomical circuits underlying these rsFC changes. Using a cross-species brain parcellation, we identified in non-human primates locations homologous to the regions of interest (ROIs) from studies showing pre-to-post-neurofeedback changes in rsFC with the left amygdala. We injected bidirectional tracers in the basolateral, lateral, and central amygdala nuclei of adult macaques and used bright- and dark-field microscopy to identify cells and axon terminals in each ROI (SN: anterior cingulate, ventrolateral, and insular cortices; DMN: temporal pole, middle frontal gyrus, angular gyrus, precuneus, posterior cingulate cortex, parahippocampal gyrus, hippocampus, and thalamus). We also performed additional injections in specific ROIs to validate the results following amygdala injections and delineate potential disynaptic pathways. Finally, we used high-resolution diffusion MRI data from four post-mortem macaque brains and one in vivo human brain to translate our findings to the neuroimaging domain. Different amygdala nuclei had significant monosynaptic connections with all the SN and DMN ipsilateral ROIs. Amygdala connections with the DMN contralateral ROIs are disynaptic through the hippocampus and parahippocampal gyrus. Diffusion MRI in both species benefitted from using the ground-truth tracer data to validate its findings, as we identified false-negative ipsilateral and false-positive contralateral connectivity results. This study provides the foundation for future causal investigations of amygdala neurofeedback modulation of the SN and DMN through these anatomic connections.

Fig. 1. Amygdala connections with the ipsilateral SN nodes.

A.i Red circles indicate the peak location of the rsFC changes after amygdala neurofeedback for the dACC in the human MNI template and the homologous location in the macaque F99 template. A.ii 3D models represent the rostro-caudal location of coronal slices from each node. A.iii The square box shows the injection site for two representative cases (in different animals), and the schematic coronal sections highlight the location with connectivity chartings in red. Individual cells are shown as red dots, dense/moderate terminals as light blue shaded areas, and diffuse terminals as light green shaded areas. A.iv Connectivity strengths for each case are shown in bar plots, with the first y-axis (red) representing the strength of the ROI inputs to the amygdala and the second y-axis (blue) representing the strengths of amygdala outputs to the ROI. *Case 7 had only anterograde transport and no input strength value. Additional cases are shown in Supplementary Fig. 2. The same organization followed for ROIs in the AI B and LPFC C. BL basolateral nucleus, BM basomedial nucleus, C caudal, Ce central nucleus, La lateral nucleus, R rostral.

Fig. 2. Amygdala connections with the ipsilateral DMN nodes.

The same organization of Fig. 1 is used to show amygdala connections (representative cases from different animals) with the Middle Frontal Gyrus A, Temporal Pole B, Parahippocampal Gyrus C, Lateral Precuneus D, Medial Precuneus E and Angular Gyrus F. Additional cases are shown in Supplementary Fig. 3. BL basolateral nucleus, BM basomedial nucleus, C caudal, Ce central nucleus, La lateral nucleus, R rostral.

Fig. 2. Amygdala connections with the ipsilateral DMN nodes.

The same organization of Fig. 1 is used to show amygdala connections (representative cases from different animals) with the Middle Frontal Gyrus A, Temporal Pole B, Parahippocampal Gyrus C, Lateral Precuneus D, Medial Precuneus E and Angular Gyrus F. Additional cases are shown in Supplementary Fig. 3. BL basolateral nucleus, BM basomedial nucleus, C caudal, Ce central nucleus, La lateral nucleus, R rostral.

Fig. 3. Validation of amygdala connections after cortical injections.

A 3D representation of the three rostro-caudal levels (i-iii) used to chart cells and terminals in the amygdala, and the respective coronal slices with cytoarchitectonic divisions based on the Paxinos atlas [38]. Labeling of cells (red dots), and dense/moderate (light blue) and diffuse (light green) terminal fields in the amygdala after bidirectional tracer injections (in different animals) in regions homologous to the Anterior Insula B and Lateral Precuneus C regions with resting-state functional connectivity changes after amygdala neurofeedback. AA anterior amygdaloid area, BL basolateral nucleus, BM basomedial nucleus, C caudal, Ce central nucleus, La lateral nucleus, R rostral.

Fig. 4. Identification of amygdala connections using dMRI tractography in NHPs and humans.

Reconstruction of streamlines (yellow) connecting the amygdala (red) with all ipsilateral nodes (green) within the SN and DMN. For each node, results from one NHP brain are shown on the left, and results from the human brain on the right.

Fig. 5. Summary of anatomical connections modulated by neurofeedback of the left amygdala.

Representation of the specificity of ROIs in the left hemisphere (red circles) overlapping major functional regions (colored parcellation). Blue and pink arrows represent monosynaptic connections from the amygdala to the SN and DMN ipsilateral ROIs, respectively. Green arrows show the disynaptic connections with the DMN contralateral ROIs through the hippocampus and PHG.