different mechanisms and temperature

NMODL and the Channel Builder.
Post Reply
menica
Posts: 68
Joined: Wed Sep 02, 2015 11:02 am

different mechanisms and temperature

Post by menica » Fri Apr 08, 2016 9:24 am

Dear Ted,
I have a two compartments model. In one of this compartments I inserted two temperature-dependent mechanisms: hh.mod and Na/K pump.
The hh is the classical built in NEURON model. In this model the parameters are setted at 6.3 degC.
The parameters of the Na/K pump model are setted at the experimental temperature of 36 degC.
Both the two mechanisms have the q10 factor, so they would adapt to any temperature choosen in the hoc code.
I noticed that if I run the silulation at celsius=6.3, without stimuli, the membrane potential stays at its rest value.
If I run it at celsius=35 the membrane potential droops at a more negative value.
To avoid this should I increase the leakage current? Would it be reasonable choose differents hh parameters values for the two compartments?
Could be the problem in using models setted at different experimental temperature (even if there is the q10 factor)?

Best Regards
Menica

ted
Site Admin
Posts: 5266
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: different mechanisms and temperature

Post by ted » Mon Apr 11, 2016 11:56 am

Most channel models ignore temperature effects. Those that try to incorporate them are generally based on little or no experimental evidence about that particular channel, but merely assume a Q10 of 3 (despite considerable evidence that many channels have Q10s that are significantly different from 3). So at best most channel models that include temperature effects employ rough first-order approximations. And some channel models specified in NMODL include a "channel density fudge factor" that makes the effective channel density a function of temperature--copying a hack that Mainen and Sejnowski used in the source code for their excellent paper
Mainen ZF, Sejnowski TJ (1996)
Influence of dendritic structure on firing pattern in model neocortical neurons.
Nature 382:363-6
See the comments about this in their na.mod and other relevant mod files.
menica wrote:Both the two mechanisms have the q10 factor, so they would adapt to any temperature choosen in the hoc code.
It is most unlikely that either mechanism's q10 factor is based on solid experimental evidence. There is no guarantee that either will work "correctly" (in the sense of doing what the biological original would do) if the model's operating temperature strays far from the temperature at which the mechanism was originally experimentally characterized. So the active transport mechanism is probaby OK at 35 deg C, but hh.mod? Not likely. Hodgkin and Huxley didn't push temperature up that high in their original papers.
I noticed that if I run the stimulation at celsius=6.3, without stimuli, the membrane potential stays at its rest value.
If I run it at celsius=35 the membrane potential droops at a more negative value.
Not a surprise, really. At 6.3 deg C, the pump is running at about 1/27 of the rate it would at 36 deg C, so it won't generate much ion flux or net current. If you want to avoid hyperpolarization at higher temperatures, do what an experimentalist would do: inject a constant current that prevents membrane potential from changing. This is far superior to changing any ionic conductance or channel gating properties, because those changes would introduce a confounding effect (if you compare results at two temperatures you wouldn't know whether to attribute an observed effect to the parameter changes you made, or to the effect of temperature).

An experimentalist would do this by injecting a "bias current" through a microelectrode, and adjusting it to force resting membrane potential to remain the same at any temperature--which means that resting membrane potential would only be controlled at that location. A modeler has the abiltity to inject a different bias current into each compartment and adjust each compartment's current automatically during initialization. See Initializing to a particular resting potential in chapter 8 of The NEURON Book, which illustrates the use of a density mechanism with the suffix constant. In a multicompartment model, one would
forall insert constant
after model setup, and the custom proc init() would be

Code: Select all

proc init() {
  finitialize(v_init)
  forall for (x,0) ic_constant(x) = -(ina + ik + il_hh + i_pump)
  if (cvode.active()) {
    cvode.re_init()
  } else {
    fcurrent()
  }
  frecord_init()
}
If you don't have The NEURON Book, pick up this prepublication draft of chapter 8
https://www.neuron.yale.edu/ftp/ted/boo ... xedref.pdf

menica
Posts: 68
Joined: Wed Sep 02, 2015 11:02 am

Re: different mechanisms and temperature

Post by menica » Wed Apr 13, 2016 6:55 am

Dear Ted,
thanks for your answer.
Looking at the code used in:
Mainen ZF, Sejnowski TJ (1996)
Influence of dendritic structure on firing pattern in model neocortical neurons.
Nature 382:363-6
It is a model suitable for neocortical dendtrite.
I was wandering if I could use these na.mod and kv.mod instead of hh.mod.
Would I resolve the problem without insert a bias current?

Reagding the init() procedure:
It is the same if I insert a Iclamp point process for one compartment?

ted
Site Admin
Posts: 5266
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: different mechanisms and temperature

Post by ted » Wed Apr 13, 2016 9:31 am

menica wrote:I was wandering if I could use these na.mod and kv.mod instead of hh.mod.
Would I resolve the problem without insert a bias current?
No, that would only create new problems. Their mechanisms were intended to be used at only one temperature.
Reagding the init() procedure:
It is the same if I insert a Iclamp point process for one compartment?
It is "the same" only in concept. Implementation would be quite different. For one thing, the current delivered by an IClamp would have to be scaled by compartment area. For another, attaching an IClamp to only one compartment would only allow controlling the membrane potential at t=0 in that compartment--any other compartment that has a different set of membrane currents would probably have its own preferred resting potential, depending on the conductance densities of whatever voltage-gated channels were present, or the existence of any other generators of membrane current (e.g. electrogenic pumps). Consequently, each simulation would begin with an interval in which membrane potential starts out at v_init everywhere at t=0 but drifts away from that over time until the model settles to a steady state, in which v will be nonuniform. That would leave you with the same confound you started with: being unable to separate the effects of temperature on resting potential from the effects of temperature on channel kinetics.

For a model that has nonuniform membrane properties, ensuring that resting potential is independent of temperature requires adjustment of the bias current delivered by iconstant in each compartment of every section, and that requires taking the varying properties of sections into account. Not all membrane current sources WRITE ina or ik--some WRITE ica, and others generate a nonspecific current. proc init() must contain a set of conditional statements that take these differences into account. One way to do this is to use ismembrane() to test for the existence of particular density mechanism, e.g. if all sections have mechanisms that WRITE ina and ik but some also contain mechanisms that WRITE ica, one could do this:

Code: Select all

if (ismembrane("ca_ion")) {
  forall for (x,0) ic_constant(x) = -(ina + ik + il_hh + ica)
} else {
  forall for (x,0) ic_constant(x) = -(ina + ik + il_hh)
}

menica
Posts: 68
Joined: Wed Sep 02, 2015 11:02 am

Re: different mechanisms and temperature

Post by menica » Wed Apr 13, 2016 1:42 pm

Dear Ted,
thanks, it is clear.
Last question to see if I understood correctly.
If I set the celsius=37, let's say, in the hoc code and all the mechanisms inserted in the model are intented to be used at exactly at that temperature as kv.mod, should be not necessary use the bias current?
Otherwise, if I set the celsius=37 and there are inserted for example hh.mod and a pump mechanism which work correctly at two different temperature, I need to insert a bias current as you explained.
Moreover using hh.mod mechanism to obtain a nice AP would be better insert a Iclamp stimulus with a big amplitude to obtain a nice spike?

ted
Site Admin
Posts: 5266
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: different mechanisms and temperature

Post by ted » Thu Apr 14, 2016 12:36 pm

menica wrote:If I set the celsius=37, let's say, in the hoc code and all the mechanisms inserted in the model are intented to be used at exactly at that temperature as kv.mod, should be not necessary use the bias current?
Depends entirely on what you want to do. It makes sense to use a bias current if you want to have complete control over the resting potential in every compartment of your model, regardless of the presence or absence of any particular ion channel or ion accumulation mechanism (pump), or the values assigned to channel densities, pump rates, ionic concentrations, or ionic equilibrium potentials, or the temperature sensitivity of any mechanism or ionic equilibrium potential. As an added bonus, it ensures that your model starts in the resting state right from the beginning of the simulation.
Moreover using hh.mod mechanism to obtain a nice AP would be better insert a Iclamp stimulus with a big amplitude to obtain a nice spike?
If you want a "nice spike" at body temperatures well above 6.3 deg C, why use hh? There are plenty of published models in ModelDB whose mod files include sodium and potassium currents that generate very nice spikes at those temperatures. Some of those mod files are flawed for one reason or another (although they worked for the original authors, it would be best to avoid reusing them for new model development), but others are quite sound. You might look at recent papers by Migliore and his collaborators, but there are many other authors whose models have spike sodium and delayed rectifier potassium channels that are properly implemented. Avoid any mod file that "adjusts channel density for temperature" by doing things like this
gna = tadj*gbar*m*m*m*h
or this
gk = tadj*gbar*n
where tadj is a scale factor whose value depends on temperature. There is no biophysical basis for including such a factor in the calculation of ionic conductance.

menica
Posts: 68
Joined: Wed Sep 02, 2015 11:02 am

Re: different mechanisms and temperature

Post by menica » Fri Apr 15, 2016 6:31 am

Thanks!!

Post Reply