Modeling the recovery of bursting of the crab pyloric neurons in 2.5x [K+] saline

This repository holds the code for the project modeling the recovery of bursting of the crab pyloric neurons in 2.5x[K+] saline. In short, this is called the HighK project. The initial findings and basic background of this project is written in the Abstract section.

This project is currently divided into 4 parts (and hence 4 folders inside HighK): Network, ABPD, ABcoupledPD, and AB_multi_compartment. The Abstract is currently about ABcoupledPD. Each part has their own experiments, which are outlined in the Experiments section. Each file also explains the purpose of each file and how to run them.

"Test files" are not actually test files, but they act more like a demo on how to run the code. They also serve as making sure the actual code runs as it was intended to.

Abstract

The pyloric circuit of the stomatogastric ganglion of crustaceans is a central pattern generator that produces a motor pattern important for digestion. Recent study discovered an adaptation of the neurons in the pyloric circuit of the crab C. Borealis to an increase in extracellular potassium concentration (He et al., 2020). They showed that the pyloric dilator (PD) neuron, which is a part of the pacemaker kernel driving bursting in the network, switched to tonic spiking in response to 2.5-fold increase in extracellular [K+] (2.5x[K+]), but then recovered its bursting activity similar to baseline within 60 minutes. The recovery of bursting activity was accompanied by the change in the intrinsic excitability of PD neurons. However, the ionic mechanism of the recovery of bursting remains unknown. We generated a large family of biophysical models of a PD neuron coupled to an anterior burster (AB) neuron to study which changes in the ionic currents result in the bursting activity in 2.5x[K+]. The 2.5-fold increase in extracellular [K+] was modeled as changing the equilibrium potentials of the potassium-dependent ion currents, which are calcium-dependent potassium current (IKCA), delayed-rectifier current (IKd), and A-type potassium current (IA), from -80mV to -56mV, and the equilibrium potential of a leak current from -50mV to -38mV. We sampled sets of maximal conductances of the ionic currents of AB and PD neurons to generate two distributions that result in 1) tonically firing PD neurons in 2.5x[K+] saline (n = 12016) and 2) burst in 2.5x[K+] saline (n = 12287). We compared the sets of maximal conductances across distributions to investigate the changes necessary for switching from tonic to bursting pattern of activity. Preliminary results show a shift in the conductance distributions of multiple ionic currents, with the biggest positive shift observed in the distribution of the maximal conductances of IKCA in a PD neuron. We also observed a significant shift towards lower conductances of IKd in a PD neuron and transient calcium current (ICaT) in an AB neuron. Future investigations include adding axonal compartments to the present models to produce more biologically plausible morphology and shifting the activation curves of the ionic currents to examine other potential mechanisms of adaptation in addition to changes in the maximal conductances.

References

He, L.S., Rue, M.C.P., Morozova, E.O., Powell, D.J., James, E.J., Kar, M., Marder, E. Rapid adaptation to elevated extracellular potassium in the pyloric circuit of the crab, Cancer borealis. J Neurophysiol. 2020 May 1;123(5):2075-2089. doi: 10.1152/jn.00135.2020.

Installation

Make sure to have the following toolboxes installed in your MATLAB

1. Xolotl
2. tight_subplot

The MyXolotl folder holds custom files that need to be copied to the Xolotl toolbox for the code to work. Run initialize.sh in the MyXolotl folder and it will add the custom files to their location in Xolotl toolbox.

The core of how this project is structured

The main class is called the HKX, which is a xolotl wrapper (more info in the file). Inside contains methods that are used repeately to setup the neurons, so it will simplify the setup process.

Experiments

The generated data for these experiments are stored in this google drive folder.

Network Experiments

Exp Number	Description	Num Models
1	Simulate the prinz network while changing Ek for A, KCa, and Kd from -100mV to -40mV, but this is wrong because synaptic currents also depend on K.	150
2	Same thing as in exp1 but now also changing Echol and Eglut	150
3.1	Increasing all gbar from x2 to x8 on ABPD that switch to silent at -70mV
3.2	Using Model 221, did a grid search on gbar on ABPD ranging from x1 to x3.	6561
3.3	Same as 3.2 except simulate 10 seconds and only grab the last 5 seconds to avoid initial effect.	6561
3.4	Increase KCa on all neurons, and they become triphasic. Just to demonstrate KCa on networks.	1
4	Attach ABcPD models to the network and see what changes

ABPD Experiments

Exp Number	Description	Num Models
1	Simulate the prinz ABPD while changing Ek for A, KCa, and Kd from -100mV to -40mV.	5

ABcoupledPD Experiments

Exp Number	Search Area (gi are based on prinz)	Criteria	Num Models	Other Notes
rand_1	[0, 2*gi]	Electrically uncoupled: AB: isi > 2000ms (English: if it's a wave) PD: isi < 2000ms (English: if it's tonic) Electrically coupled: AB: isi > 2000ms (English: if it's bursting) PD: isi > 2000ms (English: if it's bursting) No restriction on frequency, *note: used diff2000, but might migrate	8895
rand_2	[0, 2*gi]	Electrically uncoupled: AB: isi > 2000ms (English: if it's a wave), firing rate: 0.5 to 1.5Hz, PD: firing rate: 3 to 20 Hz, spikes_per_burst_mean <= 1.1 Electrically coupled: AB: isi > 2000ms (English: if it's bursting) PD: isi > 2000ms (English: if it's bursting) *note: used diff2000, but might migrate	373	Sometimes it will still have 2 spikes in a burst so need to do something about it (hence rand_3)
rand_3	[0, 4*gi] for A, CaS, CaT, KCa, Kd, PD.NaV [0, 2*gi] for H and leak	Electrically uncoupled: AB: isi > 2000ms (English: if it's a wave), firing rate: 0.5 to 1.5Hz PD: firing rate: 3 to 20 Hz, spikes_per_burst_mean <= 1.1, isi_std < 200 Electrically coupled: AB: isi > 2000ms (English: if it's bursting) PD: isi > 2000ms (English: if it's bursting) *note: used diff2000, but might migrate	12016
rand_4	[0, 4*gi] for A, CaS, CaT, KCa, Kd, NaV [0, 2*gi] for H and leak *Note we added AB.NaV back	same as rand_3
rand_5	[0, 4*gi] for A, CaS, CaT, KCa, Kd, PD.NaV [0, 2*gi] for H and leak *Changed Eleak to -65mV	same as rand_3		Just want to see how different the distribution is going to be
rand_high_1	[0, 40*gi]	At -56: AB: isi > 2000ms (English: if it's bursting) PD: isi > 2000ms (English: if it's bursting)		Pilot run
rand_high_2	[0, 40*gi]	At -56: AB: isi > 2000ms (English: if it's bursting) PD: isi > 2000ms (English: if it's bursting) At -80: AB: isi > 2000ms (English: if it's bursting) PD: isi > 2000ms (English: if it's bursting)	4056
rand_high_3	[0, 80*gi]	same as rand_high_2
rand_high_4	[0, 40*gi]	At -56: AB: isi > 2000ms (English: if it's bursting) PD: isi > 2000ms (English: if it's bursting), spikes_per_burst_mean > 1 At -80: AB: isi > 2000ms (English: if it's bursting) PD: isi > 2000ms (English: if it's bursting), spikes_per_burst_mean > 1	5656
rand_high_5	[0, 20*gi]	same as rand_high_4
rand_high_6	[0, 2*gi]	same as rand_high_4	14084
rand_high_7	[0, 4*gi] for A, CaS, CaT, KCa, Kd, NaV [0, 2*gi] for H and leak	At -56: AB: Search for a good spike_threshold, then spikes_per_burst_mean >= 3, ibi_std < 100 PD: Spikes_per_burst_mean > 1, ibi_mean > 500 At -80: AB: Search for a good spike_threshold, then spikes_per_burst_mean > 1 PD: Spikes_per_burst_mean >1, ibi_mean > 500	12287
rand_high_8	[0, 4*gi] for A, CaS, CaT, KCa, Kd, NaV [0, 2*gi] for H and leak	rand_high_7 criteria with beginning of rand_3
rand_high_9	[0, 4*gi] for A, CaS, CaT, KCa, Kd, NaV [0, 2*gi] for H and leak *Added NaV in AB	same as rand_high_7
rand_high_10	[0, 4*gi] for A, CaS, CaT, KCa, Kd, NaV [0, 2*gi] for H and leak *Changed Eleak to -65mV	same as rand_high_7		Just want to see how different the distribution is going to be
rand_ac_1	[0, 4*gi] for A, CaS, CaT, KCa, Kd, NaV [0, 2*gi] for H and leak	same as rand_high_7	10053
FindCC_1	Pilot run, not sure what initial conductance values were set to		10000
FindCC_2	Set both AB and PD leak to 0.8 and NaV to 0 ABPD: [0, 100], PDAB: [0, 100], total of 100x100 = 10000 CCs		10000
FindCC_3	Set both AB and PD leak to 0.1 and NaV to 0 ABPD: [0, 100], PDAB: [0, 100], total of 100x100 = 10000 CCs		10000
FindCC_4	Set both AB and PD leak to 0.1 and NaV to 0. Search through coupling space from 1 to 100, so a total number of 10000 CCs, but this time only plot points from -4:0.25:-1.5 (because beyond -1.5 somtimes not as linear as before) to find the slope (aka CC). Also set leak to -65mV because experiment says so		10000
FindPD			256
FindABcPD-70			250
FindABcPD-56			250

AB_multi_compartment Experiments

Exp Number	Search Area	Criteria	Num Models	Other Notes
rand_1	[0, 4*gi] for A, CaS, CaT, KCa, Kd, NaV [0, 2*gi] for H and leak	At -80mV ABaxon: spikes_per_burst >= 3 ABsn: firing rate between 0.5Hz to 2Hz At -56mV, ABaxon: ibi_std < 100 (more uniform means tonic firing)	10651
rand_high_1	[0, 4*gi] for A, CaS, CaT, KCa, Kd, NaV [0, 2*gi] for H and leak	At -80mV ABaxon: spikes_per_burst >= 3 ABsn: firing rate between 0.5Hz to 2Hz At -56mV, ABaxon: ibi_mean > 500 (0.1ms), ibi_std < 2, spikes_per_burst => 3	11623