Shallow attractor basins changing the synchrony of network activity in Schizophrenia

Schizophrenia

 

Schizophrenia is a complex neurocognitive disorder, which can be roughly characterised by three kinds of symptoms – positive, negative and cognitive pointing towards the heterogeneity of the disease. Neural network modelling has been used to understand the progression of the disease conditions using EEG or MEG in Schizophrenic patients. The recent article from Yuste’s lab “ Altered cortical ensembles in mouse models of Schizophrenia” (1) had proposed that in chronic models as well as in the genetic models of Schizophrenia, the co-active ensembles possess a shallow attractor basin and they fall into one continuous distribution of activity states whilst the normal mice exhibits multiple attractor basins. This is in line with the popular notion that the recurrent excitatory collateral synapses involved in short term memory keeps the energy state low (deep attractor basins) and this would make it difficult to jump from one energy minimum to another in the normal brain, hence representing multiple activity state. In the case of schizophrenic patients, there exists a shallow basin causing spurious attractor basins which perturbs short-term memories, for instance in some cases causing hallucinations (positive symptoms of schizophrenia). NMDAR and GABAR hypo functions are the usual suspects.

The gamma oscillations generated by the local population of neurons or ensembles (both excitatory and inhibitory) is phase locked with networks in other areas of the brain thereby enabling communication between them (2). The oscillations such as gamma are entrained in a local network through attractor basins (recurrent strong connections formed between the co-active neurons), which provides strong spiking (synchronisation of activity) to transfer information from one region to another through feed forward projections. One of the surmise of having a disorganised attractor states in Schizophrenic model would be that gamma oscillations would not propagate beyond the region where it has been formed (In other words, the trafficking of communication between different regions are perturbed). These reductions in gamma power band as well as the impaired synchrony disables communication between networks. Assuming that the recurrent networks forming attractor basins strengthens the synaptic connection between the ensembles and hence the structural features of neuronal wiring determine the modification of synchronous activity between the co-active ensembles. So far it has been postulated that mostly by EEG or MEG signals that the entrained gamma oscillations do not go beyond the region of stimuli in Schizophrenic patients.

Another consequence of having a shallow attractor basin is that of a reduced signal to noise ratio. Noise is an established concept in biology that plays a pivotal role in gene expression to fate choices to neural network stability. When group of neurons or ensembles of neurons are active together over time-varying inputs, the noise of the activity in the population is co-related with the variance of the activity distributed across the neurons active for a specific task. If the attractor basins were shallow, the network activity would be noisy; this renders the network to jump from one state to another. In the case of the positive symptoms of schizophrenia such as in the state of hallucination, there is a jump of network activity (spontaneous state (stochastic firing and random fluctuations) to an active memory state, which doesn’t have any item stored in its bits or bandwidth of memory). Moreover different memories that are presented as internal representation of a particular state or perception can be shuffled without any barrier (state of free energy) in a shallow basin due to the stochastic firing of neural network. It has been proposed that networks with strong reciprocal connections are robust to noise degradation of signal.

These recent study of altered cortical ensembles in Schizophrenic mouse models (1)are conducted in the V1 region of the brain. The numbers of cells imaged were 50 to 150 cells over different days to look for the co-activation of ensembles as well as the re-activation of the ensembles. This study thereby possesses limited temporal resolution and spanning in a very small region. A few question emanates from these study are enumerated below.

1.How are ensembles activated in those models, which deviate from strict recurrent projections, for instance feed forward projections from one area of the brain to another, interlayer connectivity, recurrent network of networks?

2.How does the network are controlled to move from one dynamic state to another over time? Do they have driver nodes that control these dynamic states?

3.How well grouping of the recurrently activated ensembles, which are phase locked to the evoked stimuli can be generalised to the common modular function of cortical network and perturbed networks in Schizophrenia?

4.How does stable representation of memory is maintained in healthy states even considering the variability of firing of neurons in the network for a particular stimuli? (Sparse coding during learned trial will be one way to circumvent) How does the decrease in the stability of network due to high noise correlation is distributed across the brain of schizophrenic patients?

5.How does the plasticity at the level of structure as well as functional along with the noise plasticity changes in normal individuals as well as in schizophrenic patients chronologically in the presence or absence of specific network topology?

References:

  1. Altered cortical ensembles in mouse models of Schizophrenia, Neuron, 2017.
  2. Abnormal neural oscillations and synchrony in Schizophrenia, Nat Rev Neuroscience, 2010.

Leave a Comment