Heritability of Multivariate Gray Matter Measures in Schizophrenia

Jessica A. Turner, Vince D. Calhoun, Andrew Michael, Theo G.m. van Erp, Stefan Ehrlich, J.m. Segall, Randy L. Gollub, John G. Csernansky, Steven G. Potkin, Beng‐Choon Ho, Juan Bustillo, S. Charles Schulz, Fbirn, Lei WangTwin 2012 · DOI

Abstract

Structural brain measures are employed as endophenotypes in the search for schizophrenia susceptibility genes. We analyzed two independent structural imaging datasets with voxel-based morphometry and with source-based morphometry, a multivariate, independent components analysis, to determine the stability and heritability of regional gray matter concentration abnormalities in schizophrenia. The samples comprised 209 and 102 patients with schizophrenia and 208 and 96 healthy volunteers, respectively. The second sample additionally included non-ill siblings of participants with and without schizophrenia. A standard voxel-based analysis showed reproducible regional gray matter deficits in the affected participants compared with unrelated, unaffected controls in both datasets: patients showed significant gray matter concentration deficits in cortical frontal, temporal, and insular lobes. Source-based morphometry (SBM) was applied to the gray matter images of the entire sample to determine the effects of diagnosis on networks of covarying structures. The SBM analysis extracted 24 significant sets of covarying regions (components). Four of these components showed significantly lower gray matter concentrations in patients (p < .05). We determined the familiality of the observed SBM components based on 66 sibling pairs (25 discordant for schizophrenia). Two components, one including the medial frontal, insular, inferior frontal, and temporal lobes, and the other including the posterior occipital lobe, showed significant familiality (p < .05). We conclude that structural brain deficits in schizophrenia are replicable, and that SBM can extract unique familial and likely heritable components. SBM provides a useful data reduction technique that can provide measures that may serve as endophenotypes for schizophrenia.

Introduction

Schizophrenia is a highly heritable disorder, yet identifying the specific genes responsible has proven incredibly difficult. Researchers often look for "endophenotypes"—biological traits that are closer to the underlying genetic cause than a clinical diagnosis. While structural MRI has long shown gray matter (GM) deficits in schizophrenia, standard methods like Voxel-Based Morphometry (VBM) analyze the brain one voxel at a time. This approach often misses how different brain regions covary as functional or structural networks. For a brain measure to be a useful endophenotype, it must be both replicable across different scanners and groups, and heritable (present in non-ill relatives). This study tests whether a multivariate approach, Source-Based Morphometry (SBM), can satisfy both requirements.

Methods

The study utilized two large, independent multisite datasets: the FBIRN + MCIC dataset (209 patients, 228 controls) and the WashU-CCNM dataset (102 patients, 96 controls, and 80 siblings). Researchers compared two primary imaging analysis techniques:

  • Voxel-Based Morphometry (VBM): A standard univariate approach identifying differences in gray matter concentration at individual points.
  • Source-Based Morphometry (SBM): A multivariate approach using Independent Component Analysis (ICA) to treat the brain as a collection of overlapping networks (components) that vary together across individuals.

To determine heritability, the researchers used SOLAR (Sequential Oligogenic Linkage Analysis Routines) to analyze 66 sibling pairs, calculating familiality ($h^2$)—the portion of variance attributable to shared genetics and environment.

Results

VBM analysis confirmed standard deficits in the frontal, temporal, and insular lobes across both datasets, showing a high correlation ($r = 0.6$) between the two datasets' t-maps.

Voxel-based morphometry (VBM) results showing significant and replicable gray matter concentration deficits in patients across frontal, temporal, and insular regions.

SBM identified 24 stable components, four of which showed significantly lower gray matter in patients ($p < .05$). These components were remarkably stable across different scanner strengths (1.5T and 3.0T) and manufacturers.

Source-based morphometry (SBM) components representing networks of covarying gray matter regions. Four components (A-D) showed significant diagnostic differences.

Crucially, two of these four components were significantly heritable:

  • Component B ($h^2 = 0.43$): A network including the medial frontal, insular, inferior frontal, and temporal lobes.
  • Component C ($h^2 = 0.49$): A network primarily in the posterior occipital lobe.

Additionally, Component A (medial frontal/insular) showed a trend toward decreasing gray matter with higher lifetime doses of antipsychotic medication, highlighting the importance of accounting for treatment effects.

Discussion

Schizophrenia appears to be a systemic disruption rather than a localized lesion. SBM captures this by identifying networks of regions that shrink in tandem, providing a more robust signal than individual voxels. By testing for familiality in siblings, the authors separated structural networks tied to genetic risk from those potentially influenced by medication or illness progression.

Limitations include the relatively small sample size for heritability analysis (66 sibling pairs) and the confounding effect of antipsychotic exposure. Future longitudinal research is needed to investigate how these heritable networks change from the first episode of psychosis through chronic stages of the illness.

Summary updated

Abstract

Structural brain measures are employed as endophenotypes in the search for schizophrenia susceptibility genes. We analyzed two independent structural imaging datasets with voxel-based morphometry and with source-based morphometry, a multivariate, independent components analysis, to determine the stability and heritability of regional gray matter concentration abnormalities in schizophrenia. The samples comprised 209 and 102 patients with schizophrenia and 208 and 96 healthy volunteers, respectively. The second sample additionally included non-ill siblings of participants with and without schizophrenia. A standard voxel-based analysis showed reproducible regional gray matter deficits in the affected participants compared with unrelated, unaffected controls in both datasets: patients showed significant gray matter concentration deficits in cortical frontal, temporal, and insular lobes. Source-based morphometry (SBM) was applied to the gray matter images of the entire sample to determine the effects of diagnosis on networks of covarying structures. The SBM analysis extracted 24 significant sets of covarying regions (components). Four of these components showed significantly lower gray matter concentrations in patients (p < .05). We determined the familiality of the observed SBM components based on 66 sibling pairs (25 discordant for schizophrenia). Two components, one including the medial frontal, insular, inferior frontal, and temporal lobes, and the other including the posterior occipital lobe, showed significant familiality (p < .05). We conclude that structural brain deficits in schizophrenia are replicable, and that SBM can extract unique familial and likely heritable components. SBM provides a useful data reduction technique that can provide measures that may serve as endophenotypes for schizophrenia.
BibTeX
@inproceedings{turner2012heritability,
  author = {Jessica A. Turner and Vince D. Calhoun and Andrew Michael and Theo G.M. van Erp and Stefan Ehrlich and J.M. Segall and Randy L. Gollub and John G. Csernansky and Steven G. Potkin and Beng‐Choon Ho and Juan Bustillo and S. Charles Schulz and FBIRN and Lei Wang},
  title = {{Heritability of Multivariate Gray Matter Measures in Schizophrenia}},
  booktitle = {Twin},
  year = {2012},
  doi = {10.1017/thg.2012.1},
}
Figure