Translational Neurocircuitry

One of the unique aspects of the MPI-MR is the integration of fundamental biological research with applied research in human physiology and mechanisms of human diseases. The Translational Neurocircuitry Group complements and extends the basic research currently being pursued on metabolic processes with studies on human physiology as well as on clinical diseases and pertaining states.

Specifically, we are interested in advancing our fundamental understanding of the neural circuits that support adaptive physiological and homeostatic functions, as well as complex behaviours. Here, our goal is primarily to extend our current understanding of those biological pathways that mediate individual differences in behaviour and risk for (psycho-)pathology implicated in the (lack of) control of body weight and energy homeostasis. To this end, we explore the role of reward processing and impulsive behaviour in obesity and related disorders.

Our research strategy requires a close interaction of theoretical and experimental work, an inter-disciplinary research environment along with an infrastructure that supports prospective validation studies in humans. To facilitate this interaction, another of our interests is to improve current techniques for probing neurocircuitry in vivo. This includes diffusion and functional MRI, positron emission tomography (PET), EEG and brain stimulation techniques, such as transcutaneous vagus nerve stimulation. Over the past years, the group has built up extensive methodological and technical expertise particularly in combining structural and functional neuroimaging approaches.

Research Interests

(1) In vivo characterization of anatomical connectivity

One of the major challenges in systems neuroscience is to identify brain networks and unravel their significance for brain function. This has led to the concept of the ‘connectome’, which aims us to produce a comprehensive map of neural connections, and thereby, to better describe the physical and functional coupling among all neural elements of the brain. Great worldwide efforts are currently devoted to studying brain network structure at multiple scales: from the detailed connectivity of local neural circuits to the large-scale connection patterns among entire brain areas. We have pursued studies in both areas with a strong emphasis on techniques that are applicable in vivo.

We are specifically interested in obtaining quantitative in vivo markers of connection strength within cortico-basal ganglia circuitry and to understand how these may predict individual expressions of behaviour or disease symptoms. Here, current neuroanatomical studies indicate that these circuitries could be far more complex than previously thought and provide an important framework for understanding structural contributions to the manifestation of obesity, reward processing and impulsive behaviour. For instance, we aim to assess distinctive connectional fingerprints (Fig. 1a) that illustrate the anatomical substrate of integrated functional networks between basal ganglia and the cerebellum.

From the perspective of mapping large-scale connection patterns of the entire brain, the most promising avenue for compiling connectivity data originates from the notion that individual brain regions maintain individual connection profiles. What defines a segregated brain region is that all its structural elements share highly similar connectivity patterns and that these patterns are dissimilar between brain regions. These connectivity patterns determine the region's functional properties and also allow their anatomical delineation and mapping (Fig. 1b).

Fig. 1. Diffusion tractography-based fingerprinting is an emerging method enabling insight into individual connectional, and thus, functional parcellation of cortex and subcortical grey matter using non-invasive MR imaging in living human subjects. These methods may be applied (A) to quantify specific fibre connections in-vivo or (B) to derive connectivity-based parcellations of brain maps.

(2) Modulation of dopaminergic mechanisms underlying reward and impulsive behaviour

The contemporary view of obesity is that this condition is a complex behavioural syndrome arising from actions of differentially vulnerable individuals. This view has stimulated our methodological research towards understanding the specific mechanisms by which these hedonic behaviours interact with homeostatic control. This interaction specifically requires that we establish formal models linking the physiology of (dopaminergic) reward circuits and hormonal influence to associated expressions of behaviour. In this regard, we apply computational modelling techniques to human neuroimaging and behavioural data with the goal of obtaining quantitative in vivo markers that predict potential individual expressions of behavioural outcome. In particular, these models are individualized by including genetic and anatomical information provided by non-invasive assays for phenotyping the long-term propensity to obesity. Hence, we aim to provide a mechanistically interpretable prediction of individual outcome rather than using phenomenological categorization.

For instance, variations in the fat mass and obesity-associated (FTO) gene are currently the strongest known genetic factor predisposing humans to non-monogenic obesity. Recent experiments have linked these variants to a broad spectrum of behavioural alterations including food choice and substance abuse. Given that Fto regulates D2/3R signalling in mice, we tested in humans whether variants in FTO would interact with a variant in the ANKK1 gene, which in turn alters D2R signalling and is also associated with obesity as well. In a combined behavioural and fMRI study, we demonstrated that gene variants of FTO affect DA (D2)-dependent midbrain responses to reward learning and behavioural responses associated with learning from negative outcome in humans. Furthermore, dynamic causal modelling confirmed that FTO variants modulate the connectivity in a basic reward circuit of meso-striato-prefrontal regions (Fig. 2), thereby suggesting a mechanism by which genetic predisposition alters reward processing not only in obesity, but also in other disorders with altered D2R-dependent impulse control, such as addiction.

Fig. 2. Variants of FTO affect dopamine-dependent midbrain responses and learning from negative outcomes in humans during a reward learning task. Specifically, FTO variants modulate the connectivity in a basic reward circuit of meso-striato-prefrontal regions (A), where connectional differences are apparent in both brain function (B) and structure (C) already in non-obese individuals, thereby contributing to a mechanistic understanding of why FTO is a predisposing factor for obesity.

(3) Brain networks implicated in insulin signalling

Insulin signalling in the central nervous system (CNS) is crucial for the regulation of energy as well as glucose homeostasis. In the case of type 2 diabetes mellitus (T2D), insulin resistance in different brain regions can impair energy homeostasis and augment disease progression. In approaches combining highly resolved diffusion MR images to compute connectional fingerprints and infer differences in connectional network aspects associated with, e.g., gene status, fMRI and pharmacological/hormonal intervention, we are specifically interested to address the question about how insulin modulation and its interaction with genotypes affects dopamine-receptive meso-striatal-frontal circuits and the behavioural expression of impulsivity (Fig. 3), respectively. As intervention methods we specifically explore the intra-nasal administration of insulin.

Fig. 3. Investigating insulin modulation of dopaminergic mechanisms underlying impulsivity. Generative modelling techniques allow to obtain quantitative in vivo markers of dopaminergic transmission and its modulation by hormones, and thus, furnish a mechanistic understanding of insulin action in the CNS.

(4) Approaches to computational phenotyping and individualized prediction

The integration of imaging information with non-invasive assays for phenotyping into computational modelling approaches can be used as a valuable tool for understanding and interpreting clinical data. In this context, we explore the perspective of an individualized prediction of disease progression and, as a consequence thereof, personalized sustainable therapeutic approaches. From a methodological point of view, we have followed complementary strategies in several different projects over the past years: both data and theory driven as well as the embedding of both.

For example, we employed hybrid modelling approaches that combine physiological and computational perspectives as well as merge state-of-the-art computational models (Hierarchical Gaussian Filters) with physiological models of effective connectivity (i.e., dynamic causal modelling) to provide estimates of individual synaptic connection strengths. Further, we included individual genetic and anatomical information. Under the premise that functional genetic polymorphisms model any emerging variability in brain chemistry, we can focus on genes that have a known or putative impact on dopaminergic processes. Computationally, this rests on a Bayesian model selection approach, which enables us to determine the relation between individual genotype and prior variances that best explain measured data.

In contrast to applications of single algorithms, we have also developed generative embedding techniques to obtain powerful and mechanistically interpretable predictions about behavioural or clinical states (Fig. 4). Here, we used machine-learning techniques to dissect spectrum disorders into clinically relevant subtypes.

Finally, we used Parkinson’s disease data as one of the most common (and DA-associated) neurodegenerative diseases to assess further methodological approaches. We introduced a multimodal approach to model symptom-side predominance in brain morphology based on multi-kernel support vector classification at an individual subject level. Especially, this research strongly advocates the use of individualized disease-modifying therapies rather than symptom-alleviating treatments.

Fig. 4. Computational assays are applied to individualized data (neuroimaging, electrophysiology, behaviour) to yield predictive models. These models may be embedded with multivariate analyses, which are utilized to estimate the optimal fit for subsequent unsupervised as well as supervised machine-learning analyses. This computational structure enables mechanistically interpretable predictions about behavioral or clinical states.

Zur Redakteursansicht