Multicompartment Soma

[claudioez]

Hi,
Using real morphology-based models with realistic conductance distributions, I would like to understand the effect of different distributions of perisomatic inhibitory synapses and the presence or absence of postsynaptic channels associated to inhibitory synapses. On those lines, I have two questions:

- Although I understand why the soma is always modelled as a single compartment, I was wondering if it makes any sense in some cases (such as mine) to model a multicompartment soma. Following that hypothetical line of work, I would also like to know if there is any estimation of the somatic conductivity (considering that in some cases the nucleus is a big part of the entire somatic area).

- In the very likely case that using a single compartment soma is the most reasonable approach, how would one model a different number of inhibitory synapses? Can one connect several synapses to the same compartment?

Thanks in advance, Claudio.

[ted]

Interesting questions.

I was wondering if it makes any sense in some cases (such as mine) to model a multicompartment soma.

Depends on the temporal and spatial scales that are relevant to the phenomena you are studying. For structures roughly the size of somas in vertebrates, there is no need to use more than one compartment for electrical signals because they equilibrate so rapidly and so nearly uniformly. Chemical signals are another matter, especially calcium (for which you may have to consider local clustering, buffering, SER, mitochondria, and also the effects of ca on second messenger cascades and ca-gated ion channels). Spatial discretization for accurate modeling of chemical signals can be implemented with NMODL or the RxD module.

I would also like to know if there is any estimation of the somatic conductivity (considering that in some cases the nucleus is a big part of the entire somatic area).

It's very difficult to determine cytoplasmic resistivity, let alone the electrical properties of organelles. A physicist might take the approach of considering a ball and stick model cell with a cylindrical dendrite and a spherical soma (you can approximate that very nicely by using pt3d syntax with 21 or more 3d data points--see https://www.neuron.yale.edu/ftp/ted/neu ... sphere.zip). Membrane properties and physical size could be based on the empirically-based ball-and-stick approximations to mammalian neurons developed back in the 1980s or so (probably CA1 hippocampal pyramidal cells, and probably published by Dan Johnston and/or Tom Brown or their colleagues). Attach a model synapse to the end of the soma opposite from where the dendrite is attached. You could use an AlphaSynapse with time constant of 3 ms and peak conductance adjusted to produce a 1 mV depolarization of the soma. Then observe the time course of membrane potential in the dendrite at its 0, 0.5, and 1 locatons. These would be the "control" results.

Next you would consider the test case: the soma contains a spherical nucleus that is a perfect insulator. I leave the implementational details of this to a future discussion.

how would one model a different number of inhibitory synapses? Can one connect several synapses to the same compartment?

You could use an AlphaSynapse with appropriate reversal potential and time course, or an Exp2Syn if it is important to describe the conductance transient with two time constants. If all are to be activated simultaneously, just scale the peak conductance of the AlphaSynapse or the weight of the NetStim that derives the Exp2Syn. If they are to be activated at different times, you'll need a separate instance for each AlphaSynapse, or if you're using Exp2Syn, just make the NetStim generate a train of events with your desired timings.

[claudioez]

Dear Ted,

thanks for such a detailed response, I will think about it and come back to you after playing a little bit.

Best, Claudio.