Application of Advanced Quantitative Methods to Schizophrenia Research - PROJECT SUMMARY
Abnormalities of white matter are important in schizophrenia. A preponderance of studies have found
decreased levels of transcripts for myelin-related proteins in autopsy brains. Some have found a decrease in the
proteins themselves, and some have not. Hundreds of diffusion tensor imaging (DTI) studies have found reduced
fractional anisotropy (FA) in the brains of many people with schizophrenia (SCH). Prefrontal white matter is
among the areas usually involved. Decreased FA is interpreted as disruption of normal architecture. However,
postmortem examination has failed to identify characteristic abnormalities, suggesting that abnormalities causing
diminished FA are subtle, and that postmortem examinations have not used the right tools to find them. We have
therefore been developing, as part of a FIC/NIMH collaboration with the Macedonian Academy of Sciences and
Arts, two new methods to characterize white matter at high resolution. The first is a machine learning protocol to
measure axonal diameters and myelin sheath thickness in electron microscope (EM) images of prefrontal white
matter, recognizing and avoiding artifacts in EM of autopsy tissue. This will enable us to measure thousands of
fibers in EM images, from individuals with SCH, major depressive disorder (MDD), or no psychiatric illness (NPI).
The second method, suggested by the DTI findings, is to analyze the spatial orientation of the axons themselves.
We will use 3-dimensional (3D) reconstructions of high-resolution images of the axons themselves, identified by
Bielschowsky silver stain or immunohistochemistry for phosphorylated neurofilament protein. To obtain high-
resolution images of Bielschowsky stains, we will take advantage of the recent observation by Dr. Mark Sonders,
co-investigator on this project, that these and other heavy metal stains luminesce under 2-photon infrared
excitation. This technique yields clear images of individual axons that can be traced and measured in 3
dimensions. We will perform these procedures on sections from existing paraffin blocks that comprise a complete
left prefrontal coronal section from 36 triads containing 1 case each of SCH, MDD, or NPI, matched for sex and
age. These brains were included in earlier studies that yielded data on protein composition, mRNA for myelin-
related proteins, DNA methylation, microglial activation, and semiquantitative myelin histology. In a third,
exploratory aim, we will employ graphical models in three multi-omics data fusion approaches to combine
different types of high-dimensional data, including those produced by Aims 1 and 2, with known structural
properties of axons and myelin in white matter, in order to build a model or detect novel dependencies of what
is disturbed in schizophrenia. We expect that novel techniques for data fusion will reveal associations based on
multidimensional correlations that could not be detected by modeling the single-domain datasets separately.