Hi!
I have been working for a while with NEURON, using Python as the interpreter, developing a biophysically realistic CA1 pyramidal cell.
At the moment, I have implemented realistic synaptic space distributions of different excitatory and inhibitory populations. The way I am doing this, instead of modelling the whole network, is by "simulating" the synaptic inputs with AlphaSynapses. For example, to account for the CA3 inputs, I distribute randomly 6209 synapses through the proximal apical dendrites, 3000 in the basal dendrites, etc.. (with a particular time distribution) and gmax, tau, etc... (precisely described in bibliography). I do exactly the same to simulate the other synaptic inputs (CA2, ECII, ECIII, AxoAxonic, PV, etc...).
I would like to record the total synaptic current in the soma that comes from each of these particular populations. For example, how much current arrives to the soma from all the scattered CA3 boutons? and from the PV boutons? Moreover, how can I track the synaptic current of one of these boutons along the neuron, until it finally arrives to the soma?
Any ideas of how should I proceed=
Thank you all in advanced!
Andrea
Tracking the synaptic current from a particular bouton from dendrite to soma

 Posts: 17
 Joined: Mon Feb 19, 2018 11:02 am
 Contact:

 Site Admin
 Posts: 5334
 Joined: Wed May 18, 2005 4:50 pm
 Location: Yale University School of Medicine
 Contact:
Re: Tracking the synaptic current from a particular bouton from dendrite to soma
For any given spatial distribution of synapses and temporal pattern of activation, a simulation will calculate the time course of the current that flows through each synapse and the current that flows through the soma membrane. This suggests a very naive approach to determining what fraction of each synapse's current reaches the soma:
Given a spatial distribution of synapses and temporal pattern of activation,
1. Run a simulation while recording the soma's transmembrane current. Call this recording i_soma_control.
2. Run another simulation while recording the soma's transmembrane current, but this time omit activation of some synapse X. Call this recording i_soma_test.
3. Subtract i_soma_test from i_soma_control. The result is an approximation to the contribution of synapse X to i_soma_control.
Why is this approach naive, and why is the result only approximate?
The fact that you are using conductance change synapses has two important consequences:
1. Synaptic activation perturbs membrane potential throughout a cell, and that affects the driving force responsible for current flow through every conductance change synapse. Consequently the amount of current and charge that enters any synapse will depend jointly on [the anatomical and biophysical properties of the cell] and [the spatiotemporal pattern of synaptic activation].
2. Activation of a conductance change synapse introduces a localized increase of conductance that will shunt some of the current that has entered through other synapses. For any particular synapse, the effect of this shunting on transfer of current and charge from synapse to soma will depend jointly on [the anatomical and biophysical properties of the cell] and [the spatiotemporal pattern of synaptic activation].
This means that the result of a single run, with a particular anatomical distribution of synapses and a particular temporal pattern of synaptic activation, represents a special case. If there are N synapses, each special case will need N+1 runs and N comparisons of test vs. control. And you will have to do that M times, where M is the number of special cases that are needed to produce an ensemble large enough for a reasonable statistical analysis of current and charge transfer from synapse to soma.
At this point you might say "Ah, but these are excitatory synapses, and each one produces only a relatively small perturbation of membrane conductance that lasts a short time (a brief conductance increase will have a localized effect because only the charge in the near vicinity of the conductance increase will have time to leak through it). Besides, there are several experimental papers that demonstrated linear spatial summation of epsps in pyramidal cells and other neurons, e.g. Nevian et al. PMID 17206140, Cash et al. PMID 9412481. Therefore we can make the assumption that shunting is minimal."
And there are several other experimental papers that demonstrated nonlinearity, e.g. Kamijo et al. PMID 25009669, Urakubo et al. PMID 15114049. Also, I should mention that Hao et al. PMID 19955407 and Li et al. PMID 25521832 published an interesting mathematical approach to describing summation of synaptic inputs, but that's a different story.
But let's suppose for a moment that spatial synaptic summation is linear. Let's also suppose that the average rate of activation of any given synapse is low so that temporal summation is linear. And finally let's assume that membrane potential is in the subthreshold range, where many kinds of cells have IV relationships that are approximately linear over a range of 10 mV or more. Then it is possible to use the results of Jaffe and Carnevale (Journal of Neurophysiology 82:3268, 1999) to do what you want, and in a very efficient way. Further discussion of this will be most efficient if we switch to email. I'll send you a note later today (Tuesday, March 13), assuming the latest snowstorm here spares electrical power and network communications.
Given a spatial distribution of synapses and temporal pattern of activation,
1. Run a simulation while recording the soma's transmembrane current. Call this recording i_soma_control.
2. Run another simulation while recording the soma's transmembrane current, but this time omit activation of some synapse X. Call this recording i_soma_test.
3. Subtract i_soma_test from i_soma_control. The result is an approximation to the contribution of synapse X to i_soma_control.
Why is this approach naive, and why is the result only approximate?
The fact that you are using conductance change synapses has two important consequences:
1. Synaptic activation perturbs membrane potential throughout a cell, and that affects the driving force responsible for current flow through every conductance change synapse. Consequently the amount of current and charge that enters any synapse will depend jointly on [the anatomical and biophysical properties of the cell] and [the spatiotemporal pattern of synaptic activation].
2. Activation of a conductance change synapse introduces a localized increase of conductance that will shunt some of the current that has entered through other synapses. For any particular synapse, the effect of this shunting on transfer of current and charge from synapse to soma will depend jointly on [the anatomical and biophysical properties of the cell] and [the spatiotemporal pattern of synaptic activation].
This means that the result of a single run, with a particular anatomical distribution of synapses and a particular temporal pattern of synaptic activation, represents a special case. If there are N synapses, each special case will need N+1 runs and N comparisons of test vs. control. And you will have to do that M times, where M is the number of special cases that are needed to produce an ensemble large enough for a reasonable statistical analysis of current and charge transfer from synapse to soma.
At this point you might say "Ah, but these are excitatory synapses, and each one produces only a relatively small perturbation of membrane conductance that lasts a short time (a brief conductance increase will have a localized effect because only the charge in the near vicinity of the conductance increase will have time to leak through it). Besides, there are several experimental papers that demonstrated linear spatial summation of epsps in pyramidal cells and other neurons, e.g. Nevian et al. PMID 17206140, Cash et al. PMID 9412481. Therefore we can make the assumption that shunting is minimal."
And there are several other experimental papers that demonstrated nonlinearity, e.g. Kamijo et al. PMID 25009669, Urakubo et al. PMID 15114049. Also, I should mention that Hao et al. PMID 19955407 and Li et al. PMID 25521832 published an interesting mathematical approach to describing summation of synaptic inputs, but that's a different story.
But let's suppose for a moment that spatial synaptic summation is linear. Let's also suppose that the average rate of activation of any given synapse is low so that temporal summation is linear. And finally let's assume that membrane potential is in the subthreshold range, where many kinds of cells have IV relationships that are approximately linear over a range of 10 mV or more. Then it is possible to use the results of Jaffe and Carnevale (Journal of Neurophysiology 82:3268, 1999) to do what you want, and in a very efficient way. Further discussion of this will be most efficient if we switch to email. I'll send you a note later today (Tuesday, March 13), assuming the latest snowstorm here spares electrical power and network communications.

 Posts: 17
 Joined: Mon Feb 19, 2018 11:02 am
 Contact:
Re: Tracking the synaptic current from a particular bouton from dendrite to soma
Thank you very much for your quick response!
I will surely look at these references, I wasn't aware of the complexity of this problem. I just assumed that there was some object/mechanism/function in NEURON that I ignored.
In fact, I am not sure of how NEURON works at lower levels. I mean, why is not possible to track how the i_bouton1 of a particular compartment is transferred to the adjacent compartment (and iteratively until reaching the soma)? If somehow this synaptic current is "felt" by the soma, how does the current (or the membrane potential change) arrive? Is it because the ionic concentrations are flowing through the compartments? Is it the difference between the membrane potential in each compartment that creates a current?
I will surely look at these references, I wasn't aware of the complexity of this problem. I just assumed that there was some object/mechanism/function in NEURON that I ignored.
In fact, I am not sure of how NEURON works at lower levels. I mean, why is not possible to track how the i_bouton1 of a particular compartment is transferred to the adjacent compartment (and iteratively until reaching the soma)? If somehow this synaptic current is "felt" by the soma, how does the current (or the membrane potential change) arrive? Is it because the ionic concentrations are flowing through the compartments? Is it the difference between the membrane potential in each compartment that creates a current?

 Site Admin
 Posts: 5334
 Joined: Wed May 18, 2005 4:50 pm
 Location: Yale University School of Medicine
 Contact:
Re: Tracking the synaptic current from a particular bouton from dendrite to soma
Before reading the rest of this particular message, it would be useful to read this articleI am not sure of how NEURON works at lower levels.
Hines, M.L. and Carnevale, N.T.
The NEURON simulation environment.
Neural Computation 9:11791209, 1997
which is available from a link on the "Publications about NEURON" page https://neuron.yale.edu/neuron/nrnpubs.
If the analogy between electrical current and the flow of water in a hydraulic system were perfect, then it would be possible for one to do the electrical equivalent of pouring a tracer into a city's reservoir and following the movement of that tracer through all the pipes as water flows out of the city's open faucets. But that's not how things work in an electrical system, not even if charge is carried by ions. If Na or Ca ions enter a cell through ligandgated channels at some synapse out on a dendrite, those ions don't zip throughout the cell. They just accumulate locally for a while, causing a local depolarization that nudges other ions in the cell to shift around a bit. In fact there will be very little movement of Na or Ca in the cell, because most intracellular mobile charge is K ions. So even if the ions that flow through the synaptic channels are "tagged" (radioactive) Na or Ca, you wouldn't be able to use that radioactivity to follow the spread of synaptic current through the cell.why is not possible to track how the i_bouton1 of a particular compartment is transferred to the adjacent compartment (and iteratively until reaching the soma)?
NEURON computes a numerical solution to the discretized cable equation (the family of ordinary differential equations that are symbolized by Equation 2 in that paper). Activation of a synapse at a particular location adds a term to the right hand side of one of those ODEs. The current that enters the cell through a synapse becomes part of the total charge in the compartment to which it is attached.
A side comment: the synaptic current loses its "identity" as soon as it has entered the cellit just adds to the total charge in the compartment that it enters. That's why you can't track it the way governments can use security cameras and your cell phone to track you. Now back to the main thread of the discussion.
Perturbation of the total charge in a particular compartment perturbs membrane potential in that compartment. The resulting perturbation spreads along the cable because of the coupling between each compartment and its immediate neighbors (that coupling is represented by the SUMMA (vk  vj)/rjk on the right hand side of Eq. 2).
hydraulic analogy slightly revised by NTC at 9:35 AM New York time on 20180314