Multiscale modeling of multitarget pharmacotherapy for dystonia

Dystonia is a movement disorder that produces involuntary sustained muscle contractions. Different types of dystonia likely involve primary or induced pathologies across multiple brain areas including basal ganglia, thalamus, cerebellum, and sensory and motor cortices. Due to lack of therapeutic alternatives, much current treatment involves paralyzing affected muscles directly with painful injections of botulinum toxin. Primary motor cortex (M1) represents a potential target for therapy. M1 pathological dynamics in some forms of dystonia include hyperexcitability and altered beta oscillations. In order to further develop understanding of motor cortex involvement in this disease and to look at potential drug cocktails (multitarget polypharmacy), we developed a multiscale model of M1 across spatial scales, ranging from molecular interactions, up to cellular and network levels. The model contains 1715 compartmental model neurons with multiple ion channels and intracellular molecular dynamics [1, 2]. Wiring and arrangements of cellular layers of the model was based on previously recorded electrophysiological data obtained from mouse M1 circuit mapping experiments. Simulations were run in the NEURON simulator and intracellular dynamics utilized the reaction–diffusion module [3]. The chemophysiological component of the simulation focused on calcium (Ca) handling, and Ca regulation of hyperpolarization-activated cyclic nucleotide-gated (HCN) channels. The Ca signaling was modeled in conjunction with intracellular cytosolic and endoplasmic reticulum (ER) volumes, inositol triphosphate (IP3) production via a metabotropic glutamate receptor signaling cascade, and ER IP3 and ryanodine receptors (RYR) which release ER Ca into the cytosol. The model reproduced the pathological dynamics providing hyperexcitability and synchronous beta oscillations across cortical layers. We applied independent random variations to multiple ion channel densities (multiple cell membrane channels: HCN, channels for Na, K, Ca; RYR, IP3 channels in ER), to identify pathological and physiological simulation sets. Experiments with these models demonstrated degeneracy, with multiple routes that produced the pathological syndrome. In most cases, there was no single parameter alteration which would induce the change from pathological to physiological dynamics. We used support vector machines to assess the high dimensional parameter space to provide overall direction for passage from an overall pathological to an overall physiological region of parameter space, enabling prediction of multitarget drug cocktails that would be likely to move the system from dystonic to physiological dynamics.