The Genetics of Structural Similarity Networks in the Brain

861 

Abstract Submission 

Isaac Sebenius¹, Varun Warrier², Richard Bethlehem³, Eva Stauffer², Richard Dear², Sarah Morgan¹, Edward Bullmore⁴

¹Cambridge University, Cambridge, Cambridgeshire, ²University of Cambridge, Cambridge, Cambridgeshire, ³Autism Research Centre, Department of Psychiatry, University of Cambridge, Cambridge, United Kingdom, ⁴University of Cambridge, Cambridge, United Kingdom

Isaac Sebenius  
Cambridge University
Cambridge, Cambridgeshire

Varun Warrier, Professor  
University of Cambridge
Cambridge, Cambridgeshire

Richard Bethlehem  
Autism Research Centre, Department of Psychiatry, University of Cambridge
Cambridge, United Kingdom

Eva Stauffer  
University of Cambridge
Cambridge, Cambridgeshire

Richard Dear  
University of Cambridge
Cambridge, Cambridgeshire

Sarah Morgan  
Cambridge University
Cambridge, Cambridgeshire

Edward Bullmore  
University of Cambridge
Cambridge, United Kingdom

Recent imaging-genetics research has demonstrated that heritable MRI-derived brain structural features show important genetic overlaps with brain function and psychopathology [1,2,3]. Yet, while the brain forms a genetically-coordinated network, existing work on the genetics of brain structure has focused on structural features at the global or regional level. As such, the genetics of network-based measures of brain structure remain largely unknown.

In this work, we conducted hundreds of genome-wide association studies (GWAS) to comprehensively characterize the genetics of structural similarity networks in the brain. Specifically, using N>30,000 subjects from the UK Biobank, we studied the genetics of Morphometric INverse Divergence (MIND), a robust and biologically-validated method to construct structural similarity networks from MRI [4]. We identified 109 independent genomic regions associated with MIND, many of which were not associated with the structural feature from which the networks were derived.

We observed positive genetic correlations between MIND network edges and the corresponding edges from functional connectivity (FC) networks, offering new evidence for a shared genetic basis for brain structure and function. Moreover, we identified putative causal relationships between MIND and functional connectivity that were specific to the association cortex.

Finally, we observed evidence for local genetic correlations between MIND network connectivity and schizophrenia, identifying specific genes such as CACNA1c that may disproportionately contribute to the shared genetic basis of brain connectivity and mental illness.

MIND networks based on neurite density index (NDI) were constructed for 31,365 subjects of European ancestry in the UK Biobank using a coarse-grained version of the HCP parcellation with 276 network edges (Fig. 1a) [4]. Using the process described in [1], we conducted a GWAS on each of these MIND network edges and the corresponding edges from FC networks. Genetic correlations were calculated using LDSC and HDL [5,6]. Mendelian randomization was performed using MRAPSS [7]. Local genetic correlations were performed with SUPERGNOVA [8].

109 independent genomic regions were associated with at least one of the 276 MIND network edges at an experiment-wide significance threshold (P < 1.8e-10, Fig. 1b). Using pairwise LDSC and Louvain clustering, we identified three genetically-defined clusters of MIND network edges corresponding to distinct anatomy and functional processes: edges corresponding to 1) visual cortex, 2) paralimbic cortex, and 3) association and somatomotor cortex (Fig. 2a).

We observed a shared genetic signal between brain structure and function; 87% of network edges showed positive genetic correlations between MIND and functional connectivity (FC), and over one-third of these relationships were statistically significant after FDR-correction (Fig. 2b). As shown in Fig. 2c, these structure-function genetic correlations were strongest in clusters 2 and 3. Using Mendelian randomization, we observed a putative causal relationship (β = 0.18, P = 0.0002) between MIND and FC specific to connectivity in cluster 3 (Fig. 2d).

We observed 81 genomic regions with significant local genetic correlations between MIND and schizophrenia (Fig. 2e). Many genomic regions showed local genetic effects in discordant directions, suggesting that global genetic correlations may not capture the full extent of the genetic relationship between brain network connectivity and SCZ. The region that most strongly contributed to MIND connectivity and SCZ fell within chromosome 12p13.33 and contained a single gene, CACNA1c, which encodes a calcium channel subunit and has been robustly associated with SCZ [9].

Through genome-wide analysis of MIND network phenotypes, we offer novel evidence for shared genetics underlying structural similarity networks, functional connectivity, and schizophrenia.

Psychiatric (eg. Depression, Anxiety, Schizophrenia)

Genetic Association Studies ¹

Connectivity (eg. functional, effective, structural) ²

Other - GWAS; MIND networks; microstructure; Schizophrenia