GIRK channels

[shyam_u2]

I want to model GIRK (G protein coupled inward rectifying potassium channel). It seems that GIRK channels are activated by G-proteins/secondary messengers (activated by metabotropic glutamate receptor) and they also seem to have Voltage dependence ( I want to model the activation/inactivation kinetics of GIRK ).Can someone suggest a methodology to model GIRK channels so that they show both Voltage dependence and G-protein coupling ?

[ted]

I want to model GIRK (G protein coupled inward rectifying potassium channel). It seems that GIRK channels are activated by G-proteins/secondary messengers (activated by metabotropic glutamate receptor) and they also seem to have Voltage dependence ( I want to model the activation/inactivation kinetics of GIRK ).Can someone suggest a methodology to model GIRK channels so that they show both Voltage dependence and G-protein coupling ?

You'll need
1. an event-driven point process that generates a G protein signal
2. a GIRK channel that is voltage- and ligand- (G protein) gated
3. something that couples 1 and 2
4. maybe a G protein accumulation mechanism, maybe not

1-3 are quite doable with a strategy similar to what I used to replicate Fig. 10 of
Gruber AJ, Solla SA, Surmeier DJ, Houk JC (2003)
Modulation of striatal single units by expected reward:
a spiny neuron model displaying dopamine-induced bistability.
J Neurophysiol 90:1095-114
You might want to examine that code--it's available from ModelDB via accession number 39949. Download the zip file, then read through readme.html on your own PC or Mac (ModelDB's own rendering of that file is mangled, so don't expect its internal links to work properly if you are reading it on-line). If that documentation and code are a bit too obscure, contact me by email
ted dot carnevale at yale dot edu
and let's work together to make something that suits your own purposes.

[shyam_u2]

Perfect. Thank you Ted. I will let you know if I face any problem..

[shyam_u2]

Can you tell a bit more about this G protein accumulation ?

[ted]

Whether it is necessary depends on the level of detail your conceptual model of what gives rise to the G protein signal. For example, it's not needed if your conceptual model is something like "a synaptic input causes a transient change in a 2nd messenger that is described by a mono- or biexponential function."

[shyam_u2]

I want subsequent spikes to produce a "summating" effect on the IPSC of post synaptic cell instead of just prolonging the effect for a specific duration. For that purpose, I inserted a piece of code as given below. Just wanna confirm whether all is good with this approach. The new line of code inserted is indicated in bold font.

NET_RECEIVE (w(1)) {

if (flag == 0) { : this is an input event
if (w>0) { : ignore events with nonpositive weight
if (msginf==1) {
msginf = 1 + w
tau = tau1
net_send(dur, 1)
} else {
msginf = msginf + w

net_move(t + dur)
}
}
} else { : this is a self-event
msginf = 1
tau = tau0
}
}

[ted]

I want subsequent spikes to produce a "summating" effect on the IPSC of post synaptic cell instead of just prolonging the effect for a specific duration.

The key question is whether your code does what you intend it to do. What is needed is an explicit and clear statement of your intent, either in unambiguous human language (ambiguity is hard to expunge) or in the form of equations and conditional statements (has an inherent clarity advantage).

With regard to temporal summation, any synaptic mechanism whose conductance is governed by one or more ODEs gives you that for free. Even ExpSyn and Exp2Syn do that, and they don't involve second messenger effects. But maybe you're not talking about summation of synaptic conductance.

If there were no "G protein" effect, how should your synaptic mechanism respond to a single activation (what is the time course of the resulting conductance change)? What should happen if it is activated repeatedly (should there be temporal summation of conductance)? Should temporal summation be linear or should it saturate, i.e. if the synapse is activated 10, 100, or 1000 times in rapid succession, should each activation cause the same increase of conductance, or should successive events in a train of activations produce a smaller and smaller response (as would occur for any real synapse that has a finite number of ligand-gated channels on the postsynaptic side)?

If there is a "G protein" effect, is it homosynaptic (i.e. caused by activation of the same presynaptic terminal that releases the transmitter that binds to the receptor associated with the postsynaptic channels)? In this case, repeated activation of presynaptic cell A would do two things:
direct action on postsynaptic ion channels
perturbation of postsynaptic G protein that in turn affects the postsynaptic ion channels

Or is it heterosynaptic (caused by activation of some other presynaptic cell)? In which case:
activation of presynaptic cell A would affect the postsynaptic ion channels but not the G protein
activation of presynaptic cell B would affect postsynaptic G protein but wouldn't by itself open the ion channels

What is the time course of the G protein effect?
If a synaptic activation occurs at t0, is there immediate potentiation of the opening of ligand-gated channels, or is the potentiating effect revealed only by a train of activations or a "paired pulse" type of experiment (conditioning pulse followed by a test pulse after some interval, in which case the conditioning pulse would elicit an unpotentiated response but the test pulse would elicit a potentiated response).
In either case, what are the equations that govern the temporal dynamics of the potentiation, does it show saturation or not, and is potentiation additive or multiplicative (does each activation add a constant amount to the G protein signal, or does it increase the G protein signal by some percentage)?

[shyam_u2]

OK. Sorry that I was not explanative. Say that I have a presynaptic neuron A which is connected to a Postsynaptic neuron B by AMPA and mglur(metabotropic glutamate receptor). Glutamate released from A binds to AMPA on neuron B and causes an EPSC. In a similar way glutamate also binds to mglur and via G protein cascade activates Kir channels to cause an IPSC in B. I am clamping neuron B to -70 mV to study its currents. I had already modeled the AMPA part by means of exp2syn point process. In order to model the effect of mglur, I am using DAsyn.mod and kir2.mod from the exmaple you had suggested. Eventually I have to tune my model according to experimental data. In the experimental data post synaptic IPSC is recorded by delivering pulses to the presynaptic neuron. If the number of pulses increase, the IPSC summates upto a certain extent and then saturates(say the IPSC summates till the number of pulses to the presynatic neuron is 10 and saturates there after).

In the example you had suggested (DAsyn.mod), subsequent events to the neuron elicits a saturating response. The first stimulus causes msg to increase to msginf at a time constant tau1 and issues a self event for a duration 'dur'. If there is a second event (within a duration of 'dur') it simply extends the presence of transmitter by shifting the self event (ie it prolongs the effect without summating). Am I right ?

In my model the post synaptic IPSC has to summate if the number of pulses to presynaptic neuron increase. In order to model this behaviour I included a piece of code as I said above which does the following. 'If the NET_RECEIVE block of DAsyn.mod receives a second stimulus within a duration of 'dur' msginf increases by a factor of synaptic weight and shifts the self event, thereby producing a summating effect on the IPSC of post synaptic cell'. Is this the right way to incorporate this behavior or do you suggest something else?

[ted]

There are two important issues that should be resolved. First is the spatial extent of the modulation of GIRK channels. Second is summation and saturation.

With regard to spatial extent: the example I suggested was intended to be illustrative, but how closely does it match your particular needs? Specifically, if a single presynaptic axon is active, what is the physical size of the region in which postsynaptic GIRK channels are affected? Is it on the order of a few microns at most? in which case it makes sense to represent GIRK channels with a point process that contains the equations that specify the I-V relationship of the channels themselves AND the equations that describe how the (very localized) G protein signal responds to a presynaptic spike. Or is the situation similar to the dopaminergic effects represented in the Gruber et al. code--either a single presynaptic axon is never active by itself, or transmitter diffuses widely over the surface of the cell so that GIRK channels are affected over a wide region? in which case the GIRK channel properties should be implemented with a density mechanism, but synaptic coupling is still represented by a point process that contains the equations that specify how presynaptic spiking generates the G protein signal.

Now we turn to summation and saturation.

In the experimental data post synaptic IPSC is recorded by delivering pulses to the presynaptic neuron. If the number of pulses increase, the IPSC summates upto a certain extent and then saturates(say the IPSC summates till the number of pulses to the presynatic neuron is 10 and saturates there after).

In the example you had suggested (DAsyn.mod), subsequent events to the neuron elicits a saturating response.

Are you saying that the Gruber et al. code saturates too "quickly"? Easily fixed--make the time constant that governs msg longer, or reduce the duration of the "transmitter pulse" shorter, or both.

It is important to not confuse "presence of transmitter" with "postsynaptic effect." Presynaptic activity at frequencies below ~1/tau1 produces postsynaptic effects that summate nearly linearly. Presynaptic activity at frequencies higher than ~1/tau1 produces postsynaptic effects that summate with saturation.

There is no need to do anything to synaptic weight. The statement
msginf = msginf + w
implements a nonsaturating additive synaptic potentiation that does not seem to be part of your conceptual model as stated in this discussion thread.

[shyam_u2]

Thank you very much. That was really helpful.

I am curious to know certain things regarding the spatial extent of G protein modulation. In my model mglur2 is located along extra synaptic dendritic membrane away from the synaptic sites so implementing GIRK channels by density mechanism and synaptic coupling by point process would be more appropriate as you had suggested.

I can understand that modeling GIRK channels by density mechanism would make sense if the transmitter spreads throughout the dendritic membrane. But how modeling by a point process (I am relatively naive about point process and stuffs) will help if the transmitter affects a small physical portion of the post synaptic membrane ?

Thank you.

[ted]

I am curious to know certain things regarding the spatial extent of G protein modulation.

So am I. I don't know whether the metGluR is located only at the postsynaptic zone, or whether there are extrajunctional receptors and significant spillover of transmitter from the synaptic cleft. Also I don't know how far the activated G protein diffuses away from the location at which it is generated inside the cell. Presumably this is in the experimental literature.

In my model mglur2 is located along extra synaptic dendritic membrane away from the synaptic sites so implementing GIRK channels by density mechanism and synaptic coupling by point process would be more appropriate as you had suggested.

So the receptors are extrajunctional and there is spillover. This implies a significant diffusional delay between transmitter release and receptor activation.

how modeling by a point process (I am relatively naive about point process and stuffs) will help if the transmitter affects a small physical portion of the post synaptic membrane ?

Point processes differ from density mechanisms in three important ways.

First, a point process is associated with a single location in space. If it is a current source, it delivers current to the segment (compartment) that contains that location. If it affects concentration of a solute, it is the concentration of that solute in the segment that contains that location. That's one reason why it makes sense to use a point process to represent conventional synaptic transmission, since the postsynaptic ion channels affected by transmitter are clustered in a region that is very small compared to the electrotonic length constants of ordinary neurites.

The second difference is purely a technical matter: point processes can respond directly to events, but density mechanisms cannot. This is why point processes are used to implement spike-triggered synaptic tranmission. For extrajunctional receptors, it may be necessary to combine a point process and a density mechanism, but only if transmitter spreads a _long_ distance from the synapse. How long? Longer than the length of the segment to which the synapse is attached. In most cases, segments are 50 um long or longer.

The third difference between point processes and density mechanisms is purely "technical, computer related stuff." One must associate a point process with a particular location in a section, but one associates a density mechanism with a section. This difference is reflected in the hoc and Python syntax for creating a point process.

Returning to the issue of spatial extent, is there evidence for glutamate action at such distances in your particular cells of interest? If not, you can get away with using a point process for your synaptic mechanism.