Computing in carbon Basic elements of neuroelectronics -- membranes -- ion channels -- wiring Elementary neuron models -- conductance based -- modelers alternatives Wires -- signal propagation -- processing in dendrites Wiring neurons together -- synapses -- long term plasticity -- short term plasticity Equivalent circuit model 1
Membrane patch The passive membrane Ohm s law: Capacitor: C = Q/V Kirchhoff: 2
Movement of ions through ion channels Energetics: qv ~ k B T V ~ 25mV The equilibrium potential K + Na +, Ca 2+ Ions move down their concentration gradient until opposed by electrostatic forces Nernst: 3
Each ion type travels through independently Different ion channels have associated conductances. A given conductance tends to move the membrane potential toward the equilibrium potential for that ion E Na ~ 50mV E Ca ~ 150mV E K ~ -80mV E Cl ~ -60mV depolarizing depolarizing hyperpolarizing shunting V E Na 0 more polarized V > E positive current will flow outward V < E positive current will flow inward V rest E K Parallel paths for ions to cross membrane Several I-V curves in parallel: New equivalent circuit: 4
Neurons are excitable Excitability arises from ion channel nonlinearity Voltage dependent transmitter dependent (synaptic) Ca dependent 5
The ion channel is a cool molecular machine K channel: open probability increases when depolarized n describes a subunit n is open probability 1 n is closed probability Transitions between states occur at voltage dependent rates P K ~ n 4 C O O C Persistent conductance Transient conductances Gate acts as in previous case Additional gate can block channel when open P Na ~ m 3 h m is activation variable h is inactivation variable m and h have opposite voltage dependences: depolarization increases m, activation hyperpolarization increases h, deinactivation 6
Dynamics of activation and inactivation We can rewrite: where Dynamics of activation and inactivation 7
Putting it together Ohm s law: and Kirchhoff s law - Capacitative current Ionic currents Externally applied current The Hodgkin-Huxley equation 8
Anatomy of a spike E K E Na Na ~ m 3 h K ~ m 3 h Anatomy of a spike E K E Na feedback Runaway +ve Double whammy 9
Where to from here? Hodgkin-Huxley Biophysical realism Molecular considerations Geometry Simplified models Analytical tractability Ion channel stochasticity 10
Microscopic models for ion channel fluctuations approach to macroscopic description Transient conductances Different from the continuous model: interdependence between inactivation and activation transitions to inactivation state 5 can occur only from 2,3 and 4 k 1, k 2, k 3 are constant, not voltage dependent 11
The integrate-and-fire neuron Like a passive membrane: but with the additional rule that when V V T, a spike is fired and V V reset. E L is the resting potential of the cell. Exponential integrate-and-fire neuron f(v) V reset V rest V th V max V f(v) = -V + exp([v-v th ]/D) 12
The theta neuron V spike V rest V th dq/dt = 1 cos q + (1+ cos q) I(t) Ermentrout and Kopell The spike response model Kernel f for subthreshold response replaces leaky integrator Kernel for spikes replaces line determine f from the linearized HH equations fit a threshold paste in the spike shape and AHP Gerstner and Kistler 13
Two-dimensional models w Simple model: V = -av + bv 2 - cw W = -dw + ev V The generalized linear model general definitions for k and h robust maximum likelihood fitting procedure Truccolo and Brown, Paninski, Pillow, Simoncelli 14
Dendritic computation Dendritic computation Dendrites as computational elements: Passive contributions to computation Active contributions to computation Examples 15
Geometry matters Injecting current I 0 r V m = I m R m Current flows uniformly out through the cell: I m = I 0 /4pr 2 Input resistance is defined as R N = V m (t )/I 0 = R m /4pr 2 Linear cables r m and r i are the membrane and axial resistances, i.e. the resistances of a thin slice of the cylinder 16
Axial and membrane resistance c m r m r i For a length L of membrane cable: r i r i L r m r m / L c m c m L The cable equation (1) x x+dx (2) 17
The cable equation (1) (2) (1) or where Time constant Space constant General solution: filter and impulse response Exponential decay Diffusive spread 18
Voltage decays exponentially away from source Current injection at x=0, T 0 Properties of passive cables Electrotonic length 19
Electrotonic length Johnson and Wu Properties of passive cables Electrotonic length Current can escape through additional pathways: speeds up decay 20
Voltage rise time Current can escape through additional pathways: speeds up decay Johnson and Wu Properties of passive cables Electrotonic length Current can escape through additional pathways: speeds up decay Cable diameter affects input resistance 21
Properties of passive cables Electrotonic length Current can escape through additional pathways: speeds up decay Cable diameter affects input resistance Cable diameter affects transmission velocity Step response: pulse travels Conduction velocity 22
Conduction velocity www.physiol.usyd.edu/au/~daved/teaching/cv.html Other factors Finite cables Active channels 23
Rall model Impedance matching: If a 3/2 = d 1 3/2 + d 2 3/2 can collapse to an equivalent cylinder with length given by electrotonic length Active cables New cable equation for each dendritic compartment 24
Who ll be my Rall model, now that my Rall model is gone Genesis, NEURON Passive computations London and Hausser, 2005 25
Enthusiastically recommended references Johnson and Wu, Foundations of Cellular Physiology, Chap 4 The classic textbook of biophysics and neurophysiology: lots of problems to work through. Good for HH, ion channels, cable theory. Koch, Biophysics of Computation Insightful compendium of ion channel contributions to neuronal computation Izhikevich, Dynamical Systems in Neuroscience An excellent primer on dynamical systems theory, applied to neuronal models Magee, Dendritic integration of excitatory synaptic input, Nature Reviews Neuroscience, 2000 Review of interesting issues in dendritic integration London and Hausser, Dendritic Computation, Annual Reviews in Neuroscience, 2005 Review of the possible computational space of dendritic processing 26