Email Icon Facebook icon Twitter Icon GooglePlus Icon Contact

User Top Menu

Neuronal Ensembles Support An “Attractive” Hypothesis for Schizophrenia

11 Apr 2017

by Hakon Heimer

It is often suggested that schizophrenia represents not a single disease process but the common final pathway for many different avenues involving a host of genetic, biological, and environmental causes. A new study led by Jordan Hamm and Rafael Yuste at Columbia University in New York City supports this notion of many paths to symptoms of schizophrenia.

In an article published online in Neuron on April 5, 2017, Hamm and colleagues write that they have found similar patterns of cortical activity abnormality in two very different models for psychosis research: a 22q11 deletion syndrome mouse model and the chronic ketamine exposure model. In both cases, the researchers found that groups of visual cortex neurons that fire together―called “ensembles”―were less consistent in this activity than corresponding ensembles in control animals.

These stable or semi-stable emergent activity patterns correspond to the “attractors” hypothesized to be the building blocks of memory, thoughts, perception, and action. Rolls and colleagues have proposed that schizophrenia is an attractor disease, wherein ensembles of cells don’t fire in harmony (Rolls et al., 2008).

“We’re suggesting that the point at which all the diverse, low-level pathophysiologies of schizophrenia converge is at the initial ensemble level in local territories of the brain,“ said Hamm. These hypothesized breakdowns of local cortical ensembles might then underlie larger-scale disruptions in connectivity seen on functional MRI.

Across models

“One of the really nice things, and one of the rare things, about this paper is that they looked across animal models,” said Philip Corlett of Yale University in New Haven, Connecticut, who was not involved in the study.

Hamm and colleagues studied visual cortex activity by several methods in two mouse models: mice that had been exposed to chronic ketamine infusions, blocking NMDA-type glutamate receptors to create a pharmacological model of schizophrenia (as reported by SRF from the 2015 SfN meeting; see also SRF glutamate hypothesis), and Df(16)A+/- mice developed by co-author Joseph Gogos to recapitulate a version of the human 22q11 deletion syndrome, which confers a high risk for schizophrenia (see SRF news here and here).

The mice were first studied with standard single-electrode electrophysiology, both at rest and during visual stimulation, and the authors report that they saw alterations in gamma-band dynamics and signal-to-noise ratio recapitulating data from human schizophrenia subjects.

The researchers then employed two-photon calcium imaging, which Yuste helped pioneer, to visually track the firing of individual and local populations of hundreds of neurons in awake animals. They found some differences between the models at the single-cell level: The chronic ketamine-treated mice, but not the Df(16)A+/- mice, had more active neurons. But those single neurons were, when averaged across trials, responding appropriately to the visual stimuli.

This was not the case at the level of local circuits. “Ensembles in both of these low-level models of schizophrenia are disrupted, both at rest and during stimulus-induced activity,” said Hamm.

Ketamine 630 Caption

While visual cortex is not a traditional hotspot for schizophrenia research, there is evidence that neurons in many different brain regions behave differently in the disorder.

“I think it’s good that we’re moving the lamp post outside of prefrontal cortex,” said Corlett. “There’s certainly evidence that it’s relevant to patients. I think about work by Dan Javitt showing early visual perturbations, particularly in first-episode patients.” (See SRF news story.)

Two control experiments did not produce the ensemble dysfunction. Both acute treatment with ketamine to block NMDA receptors and acute pharmacogenetic suppression of parvalbumin-containing interneurons produced recordable, though different, alterations at the level of single cells, yet neither manipulation produced ensemble-level changes.

“We’re providing an umbrella for all these separate findings that converge once you move from the molecular and cellular levels to the circuit level,” said Yuste.

Hamm said that they have not tried normalizing the faulty attractors with current antipsychotic drugs, though he thinks it could be informative.

"Based on the fact that first- and second-generation antipsychotic medications, acting primarily via the midbrain-striatal dopamine system, do not completely ameliorate cognitive and negative symptoms, one could expect that treatments targeting cortical circuits directly may have a more robust effect on cortical ensembles," he said. "Alternatively, one could imagine that restabilizing striatal-pallidal-thalamic dynamics via typical/atypical antipsychotics could partially recover thalamic reticular nucleus control over thalamocortical input, gradually reinforcing externally driven ensembles in a feedforward manner."

One model to explain, and treat, them all?

The authors suggest that these abnormalities could underlie different facets of schizophrenia: positive symptoms arising from inconsistent or unreliable attractors and cognitive symptoms from unstable attractors.

Corlett agrees about the potential to explain different aspects of schizophrenia, adding negative symptoms to the mix.

“I think it provides to a nice motif for both the positive symptoms of schizophrenia, wherein the network is responding to information that isn’t present, and also some of the negative symptoms, wherein it’s hard to bring the network out of what’s called a basin of attraction, to get enough energy to move to a new state,” he said.

For his part, Yuste is engaged in thought experiments beyond mere explanations. “This provides a conceptual framework to attack the pathophysiology of schizophrenia and correct it,” he said.

Last year, Yuste’s group published a study in which they used optogenetics to create artificial ensembles in mouse visual cortex, without disrupting pre-existing ones. He proposes the possibility of using optogenetics, or a related method called optochemistry, to strengthen weak attractors in patients.

Such a Star Trek medicine scenario (which Yuste and George Church sketched out in a Scientific American article; see Yuste and Church, 2014) will depend on future advances in mapping the activity of circuits to function and then identifying faulty attractors in patients.



Submitted by Edmund Rolls on

An attractor network approach to schizophrenia, and empirical tests

A computational neuroscience approach to the symptoms, mechanisms of, and treatments for schizophrenia (Loh et al., 2007; Rolls et al., 2008; Rolls and Deco, 2011; Rolls, 2012; Rolls, 2016) is supported by work on a mouse model of schizophrenia involving chronic ketamine administration (Hamm et al., 2017). The computational neuroscience approach is based on a stochastic neurodynamic framework in which the stability of attractor networks in the brain is analyzed (Rolls and Deco, 2010; Rolls, 2016). The stability is influenced by statistical fluctuations in populations of neurons caused by the neuronal spiking time randomness for a given mean firing rate. The stability of the high firing rate attractor state, which implements effects such as short-term memory and attention, is increased if the firing rates are sufficiently high to dominate the spiking-related noise. The stability of the low, spontaneous firing rate state in the absence of input must also be maintained, and GABA-mediated inhibition is important for this.

In schizophrenia, the approach (Loh et al., 2007; Rolls et al., 2008; Rolls and Deco, 2011; Rolls, 2012; Rolls, 2016) suggests that a reduction of the firing rates of cortical neurons, caused, for example, by reduced NMDA receptor function present in schizophrenia, can lead to instability of the high firing rate attractor states that normally implement short-term memory and attention in the prefrontal cortex, contributing to the cognitive symptoms of schizophrenia. Reduced NMDA receptor function in the orbitofrontal cortex by reducing firing rates may produce the negative symptoms by reducing motivation and emotion. Reduced cortical inhibition caused by a reduction of GABA neurotransmission, present in schizophrenia in, for example, the temporal lobes, can lead to instability of the spontaneous firing states of cortical networks, leading to a noise-induced jump to a high firing rate attractor state even in the absence of external inputs, contributing to the positive symptoms of schizophrenia.       

In the mouse model, two-photon calcium imaging of local cortical populations revealed a deficit in the reliability of neuronal coactivity patterns (ensembles) produced by chronic ketamine administration (Hamm et al., 2017), which blocks NMDA receptors, that is consistent with the theory of the cognitive and negative symptoms of schizophrenia, in which neuronal network activity in some cortical areas is maintained insufficiently (Loh et al., 2007; Rolls et al., 2008; Rolls and Deco, 2011; Rolls, 2012; Rolls, 2016). The reduced effects on NMDA receptors produced by the chronic NMDA receptor blockade  may also increase the activity of GABA inhibitory interneurons that may lead to increases in the activity of neurons in the hippocampal/temporal lobe-related system, which may in turn produce some of the overactivity in networks that leads to the positive symptoms of schizophrenia (Loh et al., 2007; Rolls et al., 2008; Rolls and Deco, 2011; Rolls, 2012; Rolls, 2016).

Submitted by Peter Uhlhaas on

This is a very intriguing paper that makes a number of important observations that are relevant for understanding the pathophysiology of schizophrenia. Firstly, diverse animal models converge on a dysfunction that implicates mainly neuronal ensembles rather than single-unit activity and thus highlights the importance of disturbances in large-scale neuronal dynamics as a critical component in schizophrenia. Secondly, these disturbances were brought about both by genetic risk factors as well as through blockade of NMDARs. Finally, this is one of the first preclinical studies that focuses on visual cortices that for a long time have been found to be fundamentally disturbed in schizophrenia. However, this aspect has been largely ignored in preclinical models of the disorder. Future research will need to verify in human neuroimaging data to what extent attractor dynamics are dysfunctional in schizophrenia (see commentary by Rolls) and how these observations can be linked to the important data presented in this paper.