Nens220, Lecture 11 Introduction to Realistic Neuronal Networks

1. Nens220, Lecture 11 Introduction to Realistic Neuronal Networks John Huguenard

2. Review • Synapses are dynamic • Presynaptic features • short term plasticity • facilitation and depression • Post synaptic • Ion redistribution, common for GABAA inhibition • Post translational changes, e.g. receptor phosphorylation • Neurons are dynamic • Threshold can change, based on history • Follows from HH model • Firing patterns also depend on history and other (non-HH) membrane conductances • E.g. IT, IH, gNMDA

3. Characterization of STP L4 stim, L2-3 record fields or EPSCs, 4 Hz Poisson train Varela et al 1997

4. Model incorporating facilitation and two types of depression D and F  1 = baseline Varela et al 1997

5. Postsynaptic changes do not alter the fit Similar values for depression and facilitation as in control Varela et al 1997

6. Presynaptic changes strongly influence the fit Dissimilar values for depression and facilitation compared to control. Therefore the DDF model provides a reasonable empirical fit assuming pre-synaptic modifications. Varela et al 1997

7. Synaptic Depression Abbott et al 1997

8. Depression and gain control Abbott et al 1997

9. LGN-Cortex synapse clustering Mel et al 1998

10. Orientation tuning resulting from synapse clustering Assumptions: dendritic Na and NMDA channels Mel et al 1998

11. Directional responses in visual cortex • The Livingstone model • Large pyramidal cells of Mynert may underlie directionality because of asymmetries in dendritic tree • These cells have very long dendrites (up to 1 mm) – could act as delay line. • If LGN cells were connected sequentially along dendrite then movement in one direction would result in summated dendritic potentials. Anderson et al 1999

12. Asymmetries in radial dendrites

13. Some direction selectivity is possible Best case, sequential activation along dendrite. Only obtained with very fast velocities (50-100o/sec) compared to 20o/sec recorded.

14. uIPSC properties:interneuron pyramidal neuron pair Xiang et al 2002

15. uIPSC properties:interneuron pyramidal neuron pair Xiang et al 2002

16. Expected IPSC kinetics in a L5 pyramidal neuron Xiang et al 2002

17. Thalamic oscillatory networks

18. Spatial organization of thalamic oscillations

19. Network models of thalamic oscillations

20. Effect of intra-nRt inhibition on burst timing

21. Integration and membrane time constant in behaving networks • Bernander et al (PNAS 1991), estimated that gm (and tm) could vary by a factor of 10 in simulated pyramidal neurons when incoming synaptic signals varied between 0 and 7 Hz. • Pare et al (J Neurophysiol 1998) found 10-17 times greater membrane noise in vivo than in vitro.

22. Dynamic clamp • Once a model conductance is developed it can be applied to a real neuron in a positive feedback. • State variables are integrated just as in a HH calculation, but the Vm is read from the neuron. • A new membrane current is calculated based on state variables, conductances and driving forces. • This current is injected back into the cell. • Depends on updating Im much faster than dynamics of state variables being modeled • Can be used to add or subtract a synaptic or voltage-gated conductance

23. Gain modulation Chance et al 2002

24. Separating conductance from noise Chance et al 2002

25. Making networks in Neuron New NEURON tutorial http://www.anc.ed.ac.uk/school/neuron/ Connecting cells: Objectvar syn[2] Cell[1].soma syn[0] = new ExpSyn(0) Cell[2].soma new NetCon(&Cell[1].soma.v(0.5), syn, threshold,delay, weight)

26. Conclusions • Neurons perform calculations • Simple integration is inadequate to describe I/O relationships of most neurons • Spatial inhomogeneities can give rise to microdomain calculations • Realistic models are often underconstrained • Therefore make the simplest model consistent with know features of the network.