The dysconnection hypothesis (2016)

Abstract

Twenty years have passed since the dysconnection hypothesis was first proposed (Friston and Frith, 1995; Weinberger, 1993). In that time, neuroscience has witnessed tremendous advances: we now live in a world of non-invasive neuroanatomy, computational neuroimaging and the Bayesian brain. The genomics era has come and gone. Connectomics and large-scale neuroinformatics initiatives are emerging everywhere. So where is the dysconnection hypothesis now? This article considers how the notion of schizophrenia as a dysconnection syndrome has developed - and how it has been enriched by recent advances in clinical neuroscience. In particular, we examine the dysconnection hypothesis in the context of (i) theoretical neurobiology and computational psychiatry; (ii) the empirical insights afforded by neuroimaging and associated connectomics - and (iii) how bottom-up (molecular biology and genetics) and top-down (systems biology) perspectives are converging on the mechanisms and nature of dysconnections in schizophrenia.

Keywords: Bayesian; Dysconnection; Neurogenetics; Neuromodulation; Predictive coding; Schizophrenia.

Fig. 1

predictive coding deals with the problem of inferring the causes of sparse and ambiguous sensory inputs. This is illustrated in the upper panel in terms of a shadow that can be regarded as a sensory impression. A plausible explanation for this input could be a howling canine. Predictive coding assumes that the brain has a model that generates predictions of sensory input, given a hypothesis or expectation about how that input was caused. Here, the expectation is denoted by μ and the sensory prediction it generates is summarized with g(μ). The prediction error is the difference between the input and predictions of that input. This prediction error is then used to update or revise the expectation, until prediction error is minimized. At this point, the expectation provides the best explanation or inference for the causes of sensations. Note that this inference does not have to be veridical: in the lower panel, the actual cause of sensations was a cat; however, the beholder may never know the true causes – provided that we minimize our prediction errors consistently, our model of the world will be sufficient to infer plausible causes in the outside world that are hidden behind a veil of sensations.

Fig. 2

This figure summarizes the neuronal message passing that underlies predictive coding. The basic idea is that neuronal activity encodes expectations about the causes of sensory input, where these expectations minimize prediction error. Prediction error is the difference between (ascending) sensory input and (descending) predictions of that input. This minimization rests upon recurrent neuronal interactions between different levels of cortical hierarchies. Anatomical and physiological evidence suggests that superficial pyramidal cells (grey triangles) compare the representations (at each level) with top-down predictions from deep pyramidal cells (black triangles) of higher levels. Right panel: this schematic shows a simple cortical hierarchy with ascending prediction errors and descending predictions. This graphic includes neuromodulatory gating or gain control (dotted lines) of superficial pyramidal cells that determines their relative influence on deep pyramidal cells encoding expectations (in the same level and the level above). Note that the implicit descending gain control rests on predictions of the precision of prediction errors at lower levels – and can be thought as mediating top-down attentional gain. Left panel: this provides a schematic example in the visual system: it shows the putative cells of origin of ascending or forward connections that convey prediction errors (grey arrows) and descending or backward connections that construct predictions (black arrows). The prediction errors are weighted by their expected precision, associated with projections from ventral tegmental area (VTA) and substantia nigra (SN). In this example, the frontal eye fields send predictions to primary visual cortex, which sends predictions to the lateral geniculate body. However, the frontal eye fields also send proprioceptive predictions to pontine nuclei, which are passed to the oculomotor system to cause movement through classical reflexes. Note that every top-down prediction is reciprocated with a bottom-up prediction error to ensure predictions are constrained by sensory information.

Fig. 3

Citations per year, from 1980 to 2016, when searching for TOPIC: (schizophrenia) AND (TOPIC: (disconnection) OR TOPIC: (disconnectivity) OR TOPIC: (dysconnection) OR TOPIC: (dysconnectivity)) in WEB OF SCIENCE™. The arrow indicates the first papers on the disconnection hypothesis were published.

Fig. 4

Genes associated with schizophrenia that are implicated in NMDA-receptor function and its interaction with modulatory neurotransmitter systems. Genes with high evidence – genome-wide significance or identified as the relevant gene in a common schizophrenia-associated duplication or deletion – are shown in dark green. Genes with less evidence – replication in two association and/or linkage studies in two different populations, or carrying a rare or de novo deleterious mutation in a patient – are shown in light green. The gene products and functions of these genes are listed in Table 1.