Methods

Software Simulations

Our finalized simulation workflow involved first generating different adjacency matrices that would define the connections between the four neurons in the 2 HCO circuit. Inhibitory synapses between the two neurons of each HCO were kept constant, as is consistent with the definition of HCOs involving neurons that reciprocally inhibit each other. Adjacency matrices were created for all combinations of symmetric feedforward connections (HCO1 to HCO2), as well as all symmetric feedback (HCO2 to HCO1) connections for 2 HCOs.

These matrices were input to MATLAB simulation code where they were integrated as either an Izhikevich matrix or Morris-Lecar matrix, and simulations were run with varying combinations of input parameters. The input parameter values were selected for each simulation by looping across predefined sets of discrete values for each individual parameter. For the Izhikevich model, the input parameters a, b, c, d, I_𝜇 , I_{syn_max}, and 𝜏_syn were varied, where a, b, and d describe characteristics of the recovery variable u from the Izhikevich ODE system, c represents the leak reversal potential, and I_𝜇 and I_{syn_max} represent the driving current and synaptic current, respectively, and 𝜏_syn is a time constant. For the Morris-Lecar model, the input parameters g_L, g_Ca, g_K, E_L, and I_𝜇 were varied, where g_L, g_Ca, and g_K represent leak, calcium, and potassium conductances, and E_L and I_𝜇 represent reverse potential and driving current. The voltage of each individual neuron was traced across the time course of each simulation, representing each neuron’s spiking and bursting activity.

The activity of each HCO was evaluated by whether both neurons membrane voltages were able to exceed a set threshold value; if not, the HCO was classified as silent, and otherwise as non-silent, a category which could later be further divided into various spiking behaviors such as spiking, bursting, and asymmetric. For each non-silent HCO in each simulation, period, phase, and duty cycle metrics were calculated from the two neurons and averaged for final HCO values.

Dynamic Clamp

The input and output calibration graphs (below) show the strongly linear relationship between the data on each of the axes for both the input and output calibration. In the input calibration, we have graphed the voltage input which ranges from -1V to 1V against analog to digital conversion (ADC) of the output. A similar method is used for the output calibration where the voltage input which is Vamp in this case is graphed against the digital to analog conversion (DAC) of the output.

Leech Neuron Experiments

When doing experiments with the leeches, we have three channels set up on the data acquisition (DAQ) system to view the cell’s membrane potential, the current monitor on the preamplifier, and the scaled Arduino output. Before starting the experiments each day, we upload a few test currents to the dynamic clamp with a resistor-capacitor artificial cell attached and make sure the outputs are as expected on the DAQ. We also disconnect the dynamic clamp as an input to the preamp and use the DAQ to view the regular activity from the real leech neuron to make sure it’s alive and healthy.