COMPUTATIONAL PREDICTION OF BIOCHEMICAL COMPENSATORY MECHANISMS IN SUBJECTS AT RISK OF DEVELOPING PARKINSON’S DISEASE.

Parkinson’s disease (PD) is characterised by the degeneration of substantia nigra pars compacta
dopaminergic neurons. These neurons have a highly complex axonal arborisation and a
high energy demand, so any reduction in ATP synthesis could lead to an imbalance between
demand and supply, thereby impeding normal neuronal bioenergetic requirements. The notion
of energy metabolism inevitably implicates mitochondria, the cells’ main powerhouses,
linking glycolysis to oxidative phosphorylation. In the brain, there are two types of mitochondria,
with synaptic mitochondria localised to neuronal synapses and somal mitochondria
localised to glial or neuronal somata. It has long been known that synaptic and somal
mitochondria differ in their localisation, substrate utilisation, and enzymatic activities. For
example, after biogenesis in and transport from the soma, synaptic mitochondria become
highly dependent upon oxidative phosphorylation and exhibit increased vulnerability to dysfunction
in PD, as opposed to somal mitochondria.
Since the description of the disease by the London apothecary James Parkinson in 1817
and after more than two hundred years of descriptive research, we envisaged that quantitative
computational modelling of PD will allow a cumulative, formal synthesis of the results of
this research to occur. Clearly not all at-risk subjects actually develop PD, the open question
is why? Are there biochemical compensatory mechanisms that protect some at-risk individuals
from developing PD? We addressed this question using constraint-based computational
modelling of dopaminergic neuronal metabolism, because we hypothesised that the existence
of metabolic compensatory mechanisms can be predicted using comprehensive models
of healthy, albeit at risk, and diseased dopaminergic neurons.
A systems biochemistry approach was applied to identify the metabolic pathways used by
neural models for energy generation. The mitochondrial component of an existing manual
reconstruction of human metabolism (Recon 3D) was extended with manual curation of
the biochemical literature and specialised using omics data from PD patients and controls,
to generate reconstructions of synaptic, somal, and astrocytic metabolism. Following the
imposition of experimentally-derived constraints, these reconstructions were converted into
stoichiometrically- and flux-consistent constraint-based computational models. These models
predict that PD is accompanied by a failure of the nigrostriatal glycolytic pathway and
that in silico perturbations to non-trivial reaction rates may be able to rescue this bioenergetic
phenotype. This is consistent with independent experimental reports where the enhancement
of glycolysis was shown to provide neuroprotection in PD. This is the first application of
biochemical network modelling used for the prediction of novel putative metabolic targets:
a step closer towards the treatment of idiopathic PD.