Integration across biophysical scales identifies molecular and cellular correlates of person-to-person variability in human brain connectivity

Abstract

Brain connectivity arises from interactions across biophysical scales, ranging from molecular to cellular to anatomical to network level. To date, there has been little progress toward integrated analysis across these scales. To bridge this gap, from a unique cohort of 98 individuals, we collected antemortem neuroimaging and genetic data, as well as postmortem dendritic spine morphometric, proteomic and gene expression data from the superior frontal and inferior temporal gyri. Through the integration of the molecular and dendritic spine morphology data, we identified hundreds of proteins that explain interindividual differences in functional connectivity and structural covariation. These proteins are enriched for synaptic structures and functions, energy metabolism and RNA processing. By integrating data at the genetic, molecular, subcellular and tissue levels, we link specific biochemical changes at synapses to connectivity between brain regions. These results demonstrate the feasibility of integrating data from vastly different biophysical scales to provide a more comprehensive understanding of brain connectivity.

Fig. 1. Study overview.

a, Dendritic spine morphometric, genome-wide gene expression and protein abundance data were acquired from SFG and ITG in a cohort of 98 individuals. b, Structural and functional MRI data were acquired from the same 98 individuals. c, Molecular measurements were integrated with dendritic spine morphologies to model functional connectivity and structural covariation between SFG and ITG. d, Key molecular processes and proteins associated with SFG–ITG connectivity. Bold formatting indicates associations with both functional connectivity and structural covariation.

Fig. 2. Functional and structural neuroimaging phenotypes.

a, Preprocessed functional and structural MRI data organized into a free resource for data sharing. This process involved curation (BIDS), assessment of aberrant acquisition parameters (CuBIDS), preprocessing (fMRIprep) and confound regression (XCP). b, Mean time series of hemodynamic fluctuations within brain regions from which we also measured proteomics, RNA-seq and dendritic spine morphology. c, Functional connectivity estimated by correlating all pairs of parcel time series (participant average shown), with the SFG–ITG connection being of primary interest. d, Structural MRI data were collected during the same scan sessions in which we acquired fMRI data. e, Freesurfer was applied to extract structural attributes from anatomically defined brain regions. f, Cross-correlation between structural attributes of SFG and ITG.

Fig. 3. Identification of molecular systems in SFG and ITG based on protein abundance.

a, Schematic representation of how proteins were clustered into co-abundant modules. Modules were inferred solely from proteomics data by applying a consensus clustering algorithm (SpeakEasy2 (ref. )) on the corresponding protein–protein correlation matrix, with no reference to ontology. b, SFG proteins arranged by their assigned module. Protein pairs within the same module (bounded by dashed squares) display a higher correlation than protein pairs belonging to separate modules. A short functional label was assigned to each module (‘transcription’, ‘chromatin’, etc.) based on its more highly enriched GO terms. The P value of the most enriched GO term of each module is indicated by the corresponding mauve bar (if −log₁₀(P) > 40, bars are truncated with −log₁₀(P) values stated). Modules were separately generated for SFG and ITG, with the module of interest in terms of brain connectivity highlighted in dashed red lines and red text. c, ITG counterpart of b. d, Most modules in one region have one or more highly overlapping modules in the other region (number of overlapping proteins shown in white text when −log₁₀(P) > 14), enriched for similar GO terms. ECM, extracellular matrix.

Fig. 4. Measurement of dendritic spines with protein-based functional characterization.

a, Representative ×60 bright-field image of a Golgi-stained dendrite from ITG of an exemplar participant. Scale bar = 10 µm. b, Digital 3D reconstruction of the dendritic segment performed on the bright-field image. c, Digital 3D reconstruction used for estimating spine density and morphometric attributes, including head diameter, length and volume, and assigning subclasses. Blue indicates thin spines, green indicates mushroom spines, red indicates stubby spines and yellow indicates filopodia. d, Representative zoomed-in bright-field image of a single Golgi-impregnated thin spine in the xy plane (red box from c). e, Left to right, 3D digital reconstruction of the dendrite (gray) and spine (green) in the xy plane, clockwise rotation in xyz dimensions and further rotation in xyz. f, Each spine attribute was associated against all proteins measured from the same region with enriched GO terms indicated (SFG in green and ITG in blue). g, T values of contrasts between SFG and ITG for each spine morphologic attribute are shown, with * indicating nominal differences.

Fig. 5. Associations between protein modules and brain connectivity.

a, Protein abundance of SFG synaptic module plotted against its dendritic spine fit. Each dot corresponds to an individual and the red dashed line corresponds to the linear fit between protein abundance and the dendritic spine fit values of the individuals. b, ITG counterpart of a. c, Partial contribution of each attribute toward the dendritic spine fit. d, Functional connectivity (with confounds regressed out) fitted by the dendritic spine component of SFG and ITG synaptic protein modules. Each red dot corresponds to the functional connectivity value of an individual with the red line visualizing how far it is from the fitted surface. The large curvature indicates a strong interaction effect of the modules on functional connectivity. e, SFG–ITG connection and other connections with stronger association strength between functional connectivity and the dendritic spine component of synaptic protein modules displayed. Different color dots correspond to different brain regions. In total, 3% of other connections showed higher association strength. Connections emanating from SFG or ITG are displayed in red. Other connections are displayed in yellow. f, Structural covariation counterpart of d. g, Structural covariation counterpart of e. In total, 0.4% of other connections showed higher association strength between structural covariation and the dendritic spine component of synaptic protein modules.

Fig. 6. Characterization of dendritic spine fits of proteins.

a, R² of the dendritic spine fits of all measured proteins. b, SFG proteins clustered by applying SpeakEasy2 to the partial R² of the spine attributes. Each functional label was assigned based on the most enriched GO terms of each cluster. The P values correspond to the overlap between functional connectivity-associated proteins and protein members of each cluster. c, Partial R² of each dendritic spine attribute averaged over proteins of each cluster. Partial R² profiles of SFG (green) and ITG (blue) were aligned using Hungarian clustering. d, ITG counterpart of b.

Fig. 7. Protein-connectivity associations.

a, Histogram on the number of associations found between proteins or their dendritic spine component and functional connectivity, aggregated over regions and effects tested. Red dotted line corresponds to the FDR threshold for the dendritic spine component. b, GO term enrichment based on applying GSEA to protein-connectivity association statistics. Displayed scatterplot corresponds to interaction effects. Ranking of GO terms is similar between SFG and ITG. c, Structural covariation counterpart of a. d, Structural covariation counterpart of b. e, Sankey diagram displaying enriched GO terms common between functional connectivity (FC) and structural covariation (SC) as outcome, and proteins associated with both FC and SC (in terms of main effect, interaction or joint of main and interaction effect) that overlap with members of the GO terms.