Drafting an activity dependent law operate with neural codes shaping the network structure in an ensemble – Part II

Ensemble II

      When and how does the excitatory and inhibitory neuron shape the ensemble to modulate a specific function during development? How does structural features of neuronal wiring influence or determine the emerging neuronal function, both at the level of single – neurons and of functional circuits? Quantifying the convergence and divergence of connectivity between PY to PY and PY to IN and between IN and IN (PY, Pyramidal neurons, IN, Interneurons) within the ensemble will tell us how the information is transferred as an optimal neural code. Circuit pruning, structural remodelling, local activity may change the optimality of the neural coding in different developmental stages in this ensembles. Computational models can be used to build from the structural connectivity and their neuronal properties can be studied across developmental stages, which might operate with specific neural codes in these ensembles. In essence, how network activity and network structure in the ensemble are shaped by activity dependent rules is one of the important questions to be addressed in neurosciences. However when you focus on the level of individual synapses or input plasticity at individual synapses considering each synapses as a computational unit, one of the emerging themes in molecular neurosciences is that there are intrinsic differences in cell types of excitatory neurons in terms of their molecular signatures. The extent of molecular heterogeneity by virtue of difference in gene expression profiles as a gradient along the dorsal-ventral axis of the CA1 denotes that there is subtypes of excitatory neurons in the brain whose activity might sub serve distinct functions in information processing (1). Both cell intrinsic (types of synapses, release of the neurotransmitter) and extrinsic (activity), rules such as competition, pruning, remodelling could contribute to the heterogeneity of the molecular composition of the synapses. Moreover functional changes can occur in developmental without change in the wiring. However synapse composition as well as types of synapses varies across not only in different regions in the brain (For instance, the molecules defining the PSD of the Parallel/climbing fiber /Purkinje synapse in the cerebellum would be different from cortex) but also in the different layers in the cortex. Alternatively it could be possible that there are different developmental rules that set the formation of synapses in different regions as well as in different layers in the cortex. The idea of chemo-attractant/repellents as guidance cues involved in specifying synapses and the temporal control of these cues have been documented extensively. Activity dependent mechanisms and cell intrinsic factors (For instance, trafficking of the inotropic receptors in the synapses, turnover of synaptic proteins, hebbian spike time dependent plasticity, stabilizing new spines) further refine the strength of those neuronal subtypes. Along the line, an emerging theme that recently gain attention is that the sister excitatory neurons exhibiting synchronous activity patterns while activated as well as spontaneously, initially connected electrically by gap junctions and later by chemical synapses have similar functions (2). Ko H et al had shown that recurrent synaptic connections between functionally related neurons are formed sequentially during development by reorganisation driven by activity dependent mechanisms. In an individual neuron level, different sensory inputs evoke multiple calcium hotspots on the same dendrite and individual dendritic fields represent unique set of activation patterns (3). This led to the proposition that the individual features in those neurons in a microcircuit or in a functional column or in a specialized region of the brain might make them to acts as neuronal ensembles. For instance recently, Clay Reid and colleagues (4) had shown that neurons with similar orientation preference or with same receptive fields are anatomically connected to each other (recurrent connections) and excite each other to form functional networks using the combination of large-scale EM reconstructions and two-photon calcium imaging data. These data has shown that the recurrent excitatory neurons converges their input to the proximal interneurons thereby tuning the receptive fields however the EM reconstruction of the inputs from the interneuron i.e. the divergence of the connection from the interneurons was not addressed. The interesting question which still needs to be addressed in the framework would be how sensory activity modulates the activity in thalamus and in cortex. If so, can we understand and model the internal states of the animal prior and after a sensory experience? In general in neural systems, there is a dynamic interaction between network structure and network activity as shaped by the activity dependent plasticity rules. Convergent – divergent connectivity is a crucial factor that determines the optimal neural code for information transfer between two networks. To study whether different developmental stages operate with specific neural codes is one of the important questions to be addressed in the field of circuit biology. This will eventually help to understand how functionally related neurons in specialized regions of the brain (functional areas or in individual columns) develop across developmental time window to learn on synapse formation, maturation and stabilization. I will outline some of the interesting questions surface on how this functionally assembled ensembles undergoes synapse maturation before, during and after critical period of development and how we can perturb the circuits genetically to understand the physiological properties of these neurons in the ensemble. I will also outline how these perturbations can affect network properties especially how we can model the development of networks within the ensembles and how these perturbations can glean information on the functional properties.

Key emerging questions to be tested:

  1. How does the development of projections activity dependent and independent (intra-cortical or cortico-cortical) transfer information from one region to another?
  1. What set of parameters (physiological and structural) are fine-tuned to transfer or constrain the information from one circuit to another?
  1. Will we able to draft a common activity dependent law in circuit formation? If so, how does the stability of those projections change across different developmental time windows such as before, during and after the critical period of development? What would happen to those networks during the development progression once they have been perturbed from normal development? How those circuits weigh and transfer information?
  1. Does an experimental manipulation of the circuits would help building computational models where one can implement and simulate the dynamic interaction between network structure and activity?
  1. Are there master neurons in the network, which governs the properties of the functional ensemble? How does the top-down signals in the cortex are functionally executed and what is the relevance of it in terms of its computational properties for specialised functions or regions in the cortex?

Approaches:

Imaging of activity as well as electrophysiological recordings of defined neurons during the formation of neural circuits

  1. Optical imagings of the activity of the synapses in specific micro columns in the cortex in live animals using GCaMP6 (Ca2+ imaging) during different time windows (Synapse formation, maturation and stabilisation) as well as using Arc Light GEVI, which will measure the changes in the membrane potential both in the sub-threshold levels as well as in the generation of action potential. Determining the active and passive properties by electrophysiological recordings. Reconstruction of the dendrites and axons and extracting structural parameters such as the complexity of the dendrites would help modelling multicompartmental structures of the neuronal morphology using Rall’s Cable theory (Passive properties). The properties of the ion conduction across the neuronal membrane by integrate and fire model (synaptic rise, decay, weights based on current as well as conductance) there by measuring the active properties of those neurons. To what extent does neurite excitability contribute to plasticity, input integration and computation of information in neural networks? To what extent does linear and non-linear summation happen during spiking of these neurons? Both recurrent excitatory and inhibitory inputs can be added and can be simulated using the common models such as “NEURON”.

2) Determining the active and passive properties by electrophysiological recordings. Reconstruction of the dendrites and axons and extracting structural parameters such as the complexity of the dendrites would help modelling multicompartmental structures of the neuronal morphology using Rall’s Cable theory (Passive properties). The properties of the ion conduction across the neuronal membrane by integrate and fire model (synaptic rise, decay, weights based on current as well as conductance) there by measuring the active properties of those neurons. To what extent does neurite excitability contribute to plasticity, input integration and computation of information in neural networks? To what extent does linear and non-linear summation happen during spiking of these neurons? Both recurrent excitatory and inhibitory inputs can be added and can be simulated using the common models such as “NEURON”.

3) Superimposed with functional data, Ca2+ measurements, voltage changes using GEVI, Genetically engineered voltage indicators (6), physiological recordings, LFPs, with tracing the connectome of defined regions of the brain with similar functional properties would enable to test how excitatory and inhibitory shape the spontaneous and evoked activity across different developmental time windows such as before critical period of development, during critical period as well as after the critical period of development. It is worth noting that not only cell-type specific recording is possible in freely moving animals, Mark schnitzer and colleagues has recently refined the methodological approach where one is able to record the activity of 1 million neurons in freely behaving animals encompassing different regions of the cortex (7).

 

4) Measuring spontaneous activity in the population of neurons as well as measuring the rate and temporal coding of the population of neurons during exploratory behaviours or experimental paradigms which is equivalent to input plasticity. These measurements extracted can be fed to the simulation protocols and the changes in the network can be modelled in a time- dependent manner. Moreover computational simulation experiments can be used to measure the dynamics and variability within the network as well as across the networks in time varying inputs. This would also help to understand the input-output dynamics of individual networks and how the information is transferred within the network and between the different networks. This is important, as it would tell how the network behaves or adjusts to the time varying inputs, which is a function of measuring its information processing abilities and its bandwidth of transfer. This would help learning whether the individual characteristics or the biophysical properties of the neurons make the population behave as an ensemble or the ensemble determines the properties of the distinct neuronal network as well as whether are there master neurons in the networks that shapes the properties of the functional ensembles

Changing the network architecture of the connections

  1. Changing or modifying the expression of the chemo attractants / repellants and transcription factors that are involved in regulating the laminar allocation of the neurons and confining their axons and dendrites in distinct laminas. Besides that gradients of transcription factors cause the specification of the area identities in the developing brain. Morphogens patterns the area and are further refined by the graded expression of transcription factors (Pax6, Couptf1, Tbr1, Lmo4, Bhlhb5 etc.) by the neurons. Those transcription factors not only specify areas but also the axonal projections of the neurons. For instance, Ctip1 regulates the identity of sensory areas and represses motor areas. Brain loss of Ctip1 causes the expansion of motor areas to the sensory areas. Moreover the projections of those neurons as well as the formation of sensory map are comprised. Over-expression of Fez2, in E14.5 embryos can turn the callosal projections neurons in the upper layer to corticofugal neurons and also their projections. These transcription factors play a major role in organisation of the cytoarchitectures, development of circuits, regulates the development of areas, circuits and also the projection of the neurons during development.

These transcription factors regulates a wide range of genes involved in laminar organisation, synaptogenesis, axon path finding. Manipulating transcription factors would enable us to learn what are the functional changes happens in terms of network activity both spontaneous and evoked when connections and synapse formations are perturbed in the mutant. Recordings from these neurons would help in changes in the physiological properties as well as changes in network activity in these altered mutants.

  1. Once the connectivity is changed, one can measure and simulate changing the active properties of the neurons, ion conduction across the membrane during post-synaptic potentials as well as during the back propagation of action potentials (APs). This form of dendritic excitability scales co –incident activation in different dendritic branches rendering the dendrites an alternative source of memory storage. Synaptic plasticity and co-incident detecting capability of neurons are mediated by the activation of regenerative currents primarily of Na+ and of voltage – gated Ca2+ channels in the dendritic regions leading to the generation of dendritic spikes. Dendritic spikes can detect co-incident synaptic inputs across the dendritic tree. Spatial and temporal activation of the distal dendrites with the back propagating action potential can generate bursts of somatic action potential and prior history of synaptic activity will augment this process. This form of dendritic excitability scales co –incident activation in different dendritic branches rendering the dendrites an alternative source of memory storage (8). Dendritic conductance changing the biophysical properties of the ion-channels, these channels somehow equates the summation in the different domains of the dendrite to do linear summation and the interaction of different voltage channels linearizes to spatial summation. For instance, the slow inactivating K+ channels can provide a long lasting inhibition by dampening the AP (Action potential) and EPSP (Excitatory post-synaptic potential) during high trains of excitatory activity (9). In addition, dendrites are also decorated with a hyperpolarization – activated cation channel, Ih, which operates in the resting membrane potential. The Ih channels modulate the different summation responses of the synaptic potential and blocking them reduces the threshold of back propagating AP for inducing dendritic spikes. By changing the connectivity (random and sparse) one could measure the time scale of synaptic filtering thereby playing with the biophysical properties of the gated ion channels, release probability of neurotransmitters, quantal size, and spike-frequency adaptation. This can be further adapted to neuron and synapse models where the rise and decay times of synaptic conduction, firing rates and weighed sums of individual synapses in neurons and in the population can be varied as a sum of independent poisson distribution with changing time-inputs.

p.s. The title images are taken from google images, sources: IBM blue gene models and from science daily

References:

1.Cembrowski MS et al., Spatial gene expression gradients underlie prominent heterogeneity of CA1 pyramidal neurons, Neuron, 2016.

2.Yu YC et al., Preferential electrical coupling regulated neocortical lineage – dependent microcircuit assembly, Nature, 2012.

3.Ko H et al., The emergence of functional microcircuits in visual cortex, Nature, 2013.

4.Lee WC et al., Anatomy and function of an excitatory network in the visual cortex, Nature, 2016.

5.Jia H et al., Dendritic organisation of sensory input to cortical neurons in vivo, Nature, 2010.

6.Gong Y et al., High speed recording of neural spikes in awake mice and flies with a fluorescent voltage sensor, Science, 2015.

7.Kim TH et al., Long-term optical access to an estimated one million neurons in the live mouse cortex, Cell Reports, 2016.

8.Takahashi N et al., Active cortical dendrites modulated perception, Science, 2016.

9.Johnson D et al., Dendritic potassium channels in hippocampal pyramidal neurons, J of Physiology, 2000.

 

 

Leave a Comment