The dynamics of neurological disease: integrating computational, experimental and clinical neuroscience
Prof J. R. Terry, as above.
There is a vast (and rapidly growing) amount of experimental and clinical data of the nervous system at very diverse spatial scales of activity (e.g. from sub-cellular through to whole organ), with many neurological disorders characterized by oscillations in neural activity across these disparate scales. Computer modelling and the development of associated mathematical theories provide us with a unique opportunity to integrate information from across these diverse scales of activity; leading to explanations of the potential mechanisms underlying the time-evolving dynamics and, more importantly, allowing the development of new hypotheses regarding neural function that may be tested experimentally and ultimately translated into the clinic. The purpose of this special issue is to present an overview of current integrative research in the areas of epilepsy, Parkinson’s disease and schizophrenia, where multidisciplinary relationships involving theory, experimental and clinical research are becoming increasingly established.
From the molecular mechanisms of plasticity (Kotaleski & Blackwell, 2010) through to the macroscopic electrical activity of whole brain regions (Deco et al., 2008; Coombes, 2010), there has been an explosion of interest in the development of computer models and associated mathematical theories to help understand the complexity of experimental and clinical data recordings from neural systems. There has been a growing acceptance of the crucial role played by oscillatory behaviour at these multiple scales and, more importantly, disruptions to them in a variety of neurological disorders. Thus, the purpose of this special issue is to showcase a range of recent developments that give an overview of the current state of the art within this rapidly evolving field.
First we explore the potential mechanisms that underpin oscillatory behaviour. Martell et al. (2012) follow an integrative approach combining in vitro electrophysiology in mouse neocortical slices and non-linear modelling to study the role of the N-methyl-d-aspartate (NMDA) receptor in generating multiscale oscillations at both the cellular and network level. They show that the voltage-dependent properties of the receptor directly affect both the intrinsic behaviour of the cell, as well as the relationship between the cell and the overall network, and that these interactions crucially determine the type and stability of oscillatory activity. This exciting multiscale approach to studying the nervous system is one that should be adopted more widely to enable our community to explain the ever-increasing complexity of multimodal data sets. Focussing on the macroscale, Hlinka & Coombes (2012) explore the mechanisms for the emergence of functional connectivity by highlighting the important role that local population dynamics can have in shaping emergent spatial functional connectivity patterns. A theoretical argument using a weakly coupled oscillator theory is illustrated by analysis and simulations of a model of interconnected neural populations of Wilson–Cowan type. The relation of structural to functional connectivity is studied, with a particular focus on the profound and systematic dependence of the functional connectivity patterns on the dynamics of a sub-population.
Schizophrenia is a serious neurological disorder for which altered temporal oscillations of neural activity have been shown in patients with the condition. Not only disruptions to the functional connectivity structures, but also abnormalities in gamma band synchrony have been implicated in several studies (see Uhlhass et al., 2011 for a review). In the present issue, Kömek et al. (2012) explore the role of dopamine (as a modulator of K+ conductance) as a synchronising effect within a network of excitatory and inhibitory neurons. Interestingly, findings from the computational model mirror those from clinical studies in patients with schizophrenia, where dopamine increased gamma power in patients, but decreased it in controls. This study is a further example of the concept of ‘dynamic balance’, whereby levels of excitation and inhibition interact to generate oscillations at a frequency determined by the properties of the neural network.
The vast diversity of oscillatory behaviour within the neural system is perhaps best characterized within the field of epilepsy (Richardson, 2011), where the time-scales of activity vary from the very slow (of the order of months or even years with respect to the aetiology of the disease) to the very fast (> 250 Hz, so-called high-frequency oscillations). Several different experimental modalities have implicated the role of both microscopic and macroscopic brain networks in the generation of the abnormal neural activity that is characteristic of seizure events. In the present issue, Milton (2012) considers an apparent paradox. Several recent studies have considered the existence of power laws in neuronal dynamics. These mean that dynamic phenomena are generated across every spatio-temporal scale within the brain. On the one hand these scale-free dynamics have been shown to occur in healthy neural populations in vitro (e.g. Mazzoni et al., 2007) and in vivo (e.g. Hahn et al., 2010), potentially benefiting the computational activities of the brain. On the other hand, they occur naturally in the cortex of both rats and human subjects with epilepsy (Worrell et al., 2002), and so potentially cause interference to computation. The author argues that epilepsy is a disease that results from the failure of mechanisms within the brain that confine these power law phenomena. The role of disrupted mechanisms is explored further in Wendling et al. (2012), which describes the use of multiscale computational models to interrogate experimentally recorded data to reveal mechanisms underlying oscillatory activity from inter-ictal spikes to high-frequency oscillations and seizures. Their integrative study focuses on the hippocampus, where they use experimental data obtained both in vitro (from hippocampal slice preparations) and in vivo (from both stereotactically implanted electrodes into the hippocampus and monopolar surface electrodes), as well as clinical data from a patient with mesial temporal lobe epilepsy. The reader is introduced to a number of modelling approaches, including a lumped-parameter approach (to describe the average population response) and a detailed model of the microscopic networks coupling together CA1 and CA3 cells within the hippocampus. The paper describes how different modelling approaches can both explain previously recorded phenomena as well as lead to experimentally testable predictions regarding the interplay of network structures in the creation of pathological oscillations. In a related study, Goodfellow et al. (2012a) explore the role of heterogeneity in generating the macroscale rhythms of epilepsy that are commonly recording using electroencephalography (EEG). Building upon their recent findings (Goodfellow et al., 2012b), where they demonstrated that epileptic activity could arise transiently as a consequence of connectivity structure of the tissue, in the present paper a discrete-time map is introduced to enable mathematical analysis of the role of heterogeneity. They study both spontaneous and stimulation-induced rhythms of activity, and relate these findings to rhythms recorded clinically. These ideas have particular relevance to the identification of so-called ‘seizure-onset zones’, where the use of electrical stimulation has attracted some recent interest (Valentín et al., 2005; Freestone et al., 2011). Building on a theme, the work of Blenkinsop et al. (2012) is focussed on the time evolution of macroscopic epileptic rhythms, where a neural mass model (Wendling et al., 2002) is used to describe the relationship between commonly observed wave-forms in clinical intracranial EEG recordings based upon the bifurcation structure of the generative model. Bifurcation theory is a mathematical tool for characterising the relationship between the dynamic output of a model and the parameters of the model. They then proceed to explore the consistency of evolution of multiple seizures from three different patients by tracking bifurcations during the course of each seizure. Using this method, evidence is presented for consistent mechanisms responsible for seizure evolution within each patient, where the mix of different time-scales of inhibition is suggested to play a critical role. The role of inhibition in oscillatory wave dynamics in the cortex is taken up further in Xiao et al. (2012). By experimentally probing the dynamics of rat neocortical brain slices, these authors show that γ-aminobutyric acid (GABA)ergic inhibition can increase the complexity of spatiotemporal oscillations and relate this to the signature of dynamical disease states of the brain.
Alongside Alzheimer’s disease, Parkinson’s disease is a common neurodegenerative condition, and is known to involve the loss of dopaminergic neurons in the substantia nigra. How the loss of these neurons ultimately results in the clinical symptoms of rigidity, slowness of movement and the classical tremor, and how these symptoms relate to the appearance of beta band (10–30 Hz) oscillations within the circuits of the basal ganglia remains an open question. At a more detailed level the syndrome is known to be associated with increased activity in the output nuclei of the basal ganglia, and in burst discharges, oscillatory firing and increased synchronisation of firing patterns throughout the basal ganglia, although how these relate to the motor effects of the condition remains an open question. Rubin et al. (2012) present a review covering many of these changes observed in both animal models of parkinsonism and human patients. They present new perspectives on the logic underpinning a number of computational approaches to modelling basal ganglia networks, which may enable the therapeutic effects of deep brain stimulation (DBS) to be investigated. In one such model, Nevado-Holgado et al. (2010) considered a firing rate approach, and demonstrated that the subthalamic nucleus (STN) and globus pallidus could support beta band oscillations as a consequence of their positive and negative feedback actions, in the presence of a non-oscillatory input into the system. This study of the macroscopic brain activity within this network is extended in Pavlides et al. (2012), where the authors use advanced mathematical techniques to gain an enhanced understanding of the transition to oscillations within this basal ganglia circuit based upon the synaptic transmission delays within the system. The effects of DBS are studied in Humphries & Gurney (2012), where a model of the complete basal ganglia circuitry is introduced to explore the hypothesis that a combination of basal ganglia output nuclei combining regularized firing and inhibition is a key contributor to the efficacy of STN DBS. The model provides explanatory power across a wide variety of primate STN DBS responses, and generates a number of new predictions, particularly with regards to the current ineffectiveness of optogenetic stimulation in the treatment of Parkinson’s disease. In a related study, Yousif et al. (2012) explore the spatiotemporal effects of STN DBS through the use of a multi-compartmental model of an STN neuron that can exhibit either a burst or tonic mode, and present evidence for the state-dependent impact of DBS on STN neurons, illustrating the subtlety of DBS-induced changes within the neighbourhood of the stimulating electrode. In a final article, Tsienga (this issue) considers the use of optogenetic techniques for exploring the connections between neural regions through precise manipulations of populations of neurons. The identification of different network structures is described and how these differ in health and disease are explored. There is clearly great potential to apply the techniques of optogenetics through a more specific form of stimulation, that may ultimately optimise the therapeutic benefits, whilst minimising side-effects.
In summary, mathematical and computational models provide a natural suite of tools with which to elucidate the mechanisms underpinning a variety of phenomena in neurological data recordings. Pursuing an integrative approach to the study of neural systems, combining computational, experimental and clinical research provides an exciting opportunity to enhance our understanding both of the normal functioning of the brain and crucially how disruptions to this function give rise to severe neurological conditions, such as epilepsy, schizophrenia, Alzheimer’s disease and Parkinson’s disease.
deep brain stimulation