Poster Presentation Australian & New Zealand Society of Magnetic Resonance Conference 2017

Using 13C isotopomer patterns to investigate metabolic compartmentation in brain (#88)

Lavanya B Achanta 1 2 , Benjamin D Rowlands 1 2 , Ben Cassidy 3 , Gary D Housley 1 , Caroline D Rae 1 2
  1. University of New South Wales, Randwick, NSW, Australia
  2. Neuroscience Research Australia (NeuRA), Sydney, NSW, Australia
  3. Department of Statistics, Columbia University, New York, USA

Metabolism in brain is highly compartmentalised. Many pathways cross cell and intracell boundaries with consequence exchange of metabolites. We can take advantage of 13C label and judicious choice of substrates to explore this compartmentation. We incubated Guinea pig brain cortical tissue slices with [1-13C]D-glucose and varying concentrations of either [1,2-13C]acetate or [U-13C]D-β-hydroxybutyrate (βOHB) for 90 min, extracted the slices and lyophilised the extracts and resuspended them in D2O containing EDTA and 2 mM [13C]formate as an internal reference and standard. Both [1,2-13C]acetate and [U-13C]D-β-hydroxybutyrate form [1,2-13C]acetylCo-A but the subsequent distribution of label from this compound varies significantly and also has different competitive effects on the distribution of [2-13C]acetylCo-A labelled from [1-13C]D-glucose. [1,2-13C]Acetate is metabolised mostly in the astrocytic compartment, while [U-13C]D-β-hydroxybutyrate is metabolised mostly in neuronal mitochondria. We adapted and applied a novel network analysis approach to the distribution patterns of isotopomers labelled from either [1-13C]D-glucose or [U-13C]D-β-hydroxybutyrate and [1,2-13C]acetate. We constructed conditional independence networks where the direct strength of association between each pair of metabolites is estimated using partial correlation. A standard partial correlation suffers from errors induced by the larger number of interactions to estimate from relatively few data samples. Here, we estimated partial correlations using the MISTIC algorithm [1], which regularizes the calculation by imposing a strict sparsity penalty on the estimated partial correlations, setting many directly to zero. This, in our relatively simple system, produced an interpretable metabolic network providing insights into the compartmentation of acetate and βOHB metabolism in brain.

[1] Marjanovic, G, et al. Large-scale l 0 sparse inverse covariance estimation. in Acoustics, Speech and Signal Processing (ICASSP), 2016 IEEE International Conference on. 2016: IEEE.