Enabling Discovery-Based Brain Metabolomics with Ultra-High Resolution Liquid Chromatography and Machine Learning

PROJECT SUMMARY
The overall objective of this project is to develop novel approaches to examine the dynamics of brain chemistry
and establish correlations between molecular mechanisms and differences in behavior, especially related to
cocaine use. Previous studies have established that cocaine use alters both dopamine release and neuronal
structure, contributing to compulsive drug-seeking behavior. Yet no study has comprehensively analyzed the
brain metabolome to discover additional changes in neurochemistry related to cocaine use. This knowledge gap
is primarily due to the poor resolution and sensitivity of current liquid one-dimensional chromatography-mass
spectrometry (1D-LC-MS) methods to analyze the diverse chemical composition present in the small sample
volumes of brain dialysate. This work will develop novel ultra-high resolution instrumentation and computational
methods to test our overall hypothesis that cocaine use causes both short-term and long-term metabolomic
alterations in the brain. We further hypothesize that these alterations are differentially expressed based on
behavioral phenotype and sex. In Aim 1, we will develop an in vivo metabolomic profile of the nucleus
accumbens, a brain region related to motivation and reward, by integrating miniaturized column and stationary
phase particle technology into a novel comprehensive two-dimensional LC-MS platform. The information
obtained in Aim 1 will provide a valuable resource for both our future aims and neurochemistry research. In Aim
2, we will discover the temporal and differential impacts of cocaine intake on the brain metabolome using
selectively bred high-responder (bHR) and low-responder (bLR) rats, an animal model for drug-seeking behavior.
Use of fast, ultra-high resolution capillary 1D-LC-MS and computational algorithms will efficiently discover
temporal differences in the metabolome of bHRs/bLRs and males/females. Aim 3 will characterize the differential
impact related to acquisition of cocaine self-administration on the brain metabolome using bHRs, bLRs, and
outbred rats trained to self-administer cocaine. Machine learning will demonstrate our ability to predict
compulsive drug-seeking behavior based on metabolomic differences. The information obtained from Aims 2-3
will reveal new metabolomic pathways associated with psychostimulant use, ultimately providing new targets for
early medical intervention. This F32 proposal will advance my training at the intersection of analytical chemistry,
metabolomics, and neuroscience at the research-rich environment of the University of Michigan. Included in the
training plan are the development of technical skills like microdialysis, animal models, and hypothesis-driven
research design and professional skills like mentorship, teaching pedagogy, scientific communication, and grant
preparation. Ultimately, the work outlined in this proposal will strengthen my transition to an independent research
career as a tenure-track professor.

Grant Number: 5F32DA061554-02
NIH Institute/Center: NIH
Principal Investigator: Caitlin Cain