Table discretisation for Ca dependence in KC channel

[pgleeson]

There is an issue with the table discretisation for the dependence of the KC channel conductance on [Ca2+] in the original GENESIS version of this model. See here for the script to create this channel.

More details... (Arnd, please add...)

This was discovered by Arnd Roth, Volker Steuber and Jenny Davie when converting the model from GENESIS to NEURON. This issue is illustrated by the different implementations of this channel in NEURON's NMODL format for the channel model with the incorrect table discretisation and one with a better, continuous calculation of the Ca dependence

[erikdsj]

In the original GENESIS implementation calcium activation is set by a linear table (tabchannel object). What I did not realize at the time of original implementation of the model is that because this results in 100 nM steps in the table, the activation of calcium-activated K+ channels is stepwise around threshold. It turns out to be quite difficult to replicate this behavior in NEURON, resulting in slight differences in activation of calcium-activated K+ channels in the NEURON models that affect the firing frequency of the model.

[reinoudmaex]

Thank you for the clarification. I am concerned, because a stepwise channel may very well explain the long time-constant around threshold I observed in my EJN13 paper. The way I see this, it may be that upon depolarization and CaP activation, the (Ca gate of the) KC channel is insufficiently activated to balance the CaP current.

[reinoudmaex]

I have rerun some simulations of Fig. 9 of Maex&Steuber EJN2013. The baseline Ca-concentration in Purkinje cell dendrites is about 200 nM (nanoMolar) during these network simulations. 1) For the KC channel, half-activation is 4 microM in the PC model. So even with a discretization step of 100 nM, this should be fine, because the table entries are linearly interpolated, and a MIchaelis-Menten function is almost linear far below half-activation. 2) For the K2 channel, half-activation is 200 nM so this may give a problem, although earlier simulations (in 2010) had shown that the K2 channel can be discarded (if compensated by KC). 3) To check all this I have recreated the tables with discretization steps of 10 nM. Qualitatively there is little difference, although I must admit that in this particular case the integration (permanent firing after stimulation) was somewhat less outspoken. Steps of 5 nM did not add further changes.