simulating in vivo -like synaptic input patterns in multicompartmental models

Simulating in vivo-like synaptic input patterns in multicompartmental models

• What are in vivo-like synaptic input patterns?• When are such simulations useful?• How we do it using GENESIS• Some strategies for analyzing the results

100 m

100 m

GP neuron• surface area: 17,700 m2

• number of synapses (ex/in): 1,200 / 6,800

• number of inputs / s 12,000 / 6,800

Ca3 pyramidal neuron• surface area: 38,800 m2

• number of synapses (ex/in): 17,000 / 2,000

• number of inputs / s 170,000 / 20,000

Numerical estimates of in vivo input levels

5,000 AMPA and 500 GABAA Synapses at 10 Hz

Ein = -70 mV

Eex = 0 mV

Isyn = Gin * (Vm - Ein) + Gex * (Vm - Eex)

Esyn = [(Gin*Ein) + (Gex*Eex)] / (Gin+ Gex)

Isyn = (Gin + Gex) * (Vm - Esyn)

Thousands of synapses add up to a lot of conductance!

Isyn = (300 nS) * (60-50mV) = 3 nA

(Pare D, Shink E, Gaudreau H, Destexhe A,

Lang EJ (1998). J Neurophysiol 79: 1460-70.)

High conductance state of neurons in vivo

(D. Jaeger, unpublished)

Neocortical pyramidal neurons

Striatal medium spiny neuron

Simulating in vivo-like synaptic input patterns in multicompartmental models

• What are in vivo-like synaptic input patterns?• When are such simulations useful?• How we do it using GENESIS• Some strategies for analyzing the results

Simulating in vivo-like synaptic input patterns in multicompartmental models

• When are such simulations useful? When we want to extrapolate from in vitro data to

the in vivo case– Intrinsic cell properties (ion channels, morphology)– Synaptic integration

• Temporal and spatial summation• Interactions between excitation and inhibition

When input complexity can’t be replicated in vitro– Input correlation / synchrony

(Edgerton JR, Reinhart PH (2003). J Physiol 548: 53-69.)

Small conductance K(Ca2+) channels (SK channels) regulate the firing rate of Purkinje neurons in vitro…

…but is this also true in vivo?

Effects of blocking SK channels in DCN neurons in vitro

(D. Jaeger, unpublished)

SK channel block in DCN neurons with in vivo-like background conductance levels

(D. Jaeger, unpublished)

(Destexhe A, Pare D (1999). J Neurophysiol 81: 1531-47.)

Modeled M-current (KCNQ) block with and without simulated background synaptic input

Spatial and temporal summation are reduced whenthe conductance level is high

(Destexhe A, Pare D (1999). J Neurophysiol 81: 1531-47)

(Fellous J-M et al (2003). Neuroscience 122: 811-29.)

200 msec

100 independent inputs

10 independent inputs

Input synchronization affects rate and precision

(Gauck & Jaeger, 2000.)

Simulating in vivo-like synaptic input patterns in multicompartmental models

• What are in vivo-like synaptic input patterns?• When are such simulations useful?• How we do it using GENESIS• Some strategies for analyzing the results

Steps involved in setting up the simulations

1. Cell morphology: reconstruct a filled neuron, obtain a morphology file from a colleague or the web, or make a simplified morphology model.

2. Passive parameters: Rm, Cm, Ri3. Active conductances: GENESIS tabchannel objects4. Synapse templates (AMPA, GABA, etc.):

gmax, τrise, τfall, Erev

5. Compartments: list of those receiving input6. For every independent synapse (in a loop):

1. Copy the synaptic conductance from a template library to the compt2. Create a timetable object to determine when the synapse activates3. Create a spikegen object to communicate with the synapse

Element tree structure for the simulation

Root

Cellpath Inputs

Soma

G_Na+

G_K+

AMPA

Dendrite

Dendrite

Library

AMPA synapse

G_Na+

G_K+

AMPA synapse

timetable

AMPA synapse

Soma

timetable

spikegen

spikegen

G_Na+

G_K+

1. Create synaptic conductances using synchan objects

//GENESIS script to define AMPA-type conductancefunction make_AMPA_syn

// make AMPA-type synapseif (!({exists AMPA}))

create synchan AMPAend// assign specific synapse propertiessetfield AMPA Ek {E_AMPA} setfield AMPA tau1 {tauRise_AMPA} setfield AMPA tau2 {tauFall_AMPA}

setfield AMPA gmax {G_AMPA} setfield AMPA frequency 0

end

2. Put the synaptic conductances into the library

//GENESIS script to create library template objects

//First, include my synapse and channel function filesinclude Syns.ginclude Chans.g

//Check if library already existsif (!{exists /library}) create neutral /library disable /libraryend

//Push library element, make conductance elements, pop librarypushe /library

make_AMPA_synmake_G_Namake_G_K

pope

3. For all compartments receiving input…

//Using the same random seed means you get the same timetables next time too.randseed 78923456 //Loop: for each compartment that receives a synapse… 1. copy the AMPA synapse from the library to the compartment 2. addmsg: connect the synaptic conductance to the compartment with

CHANNEL and VOLTAGE messages

//set up the timetable 1. create a unique timetable object for this compartment’s AMPA synapse 2. set timetable fields with setfield:

method: 1 = exponential distribution of intervals2 = gamma distribution of intervals3 = regular intervals4 = read times from ascii file

meth_desc1: mean interval (= 1/rate) meth_desc2: refractory period (we use 0.005) meth_desc3: order of gamma distribution (we use 3)

3. call /inputs/Excit/soma/timetable TABFILL

//set up spikegen create a unique spikegen object for this compartment’s synapse set the spikegen fields with setfield

output_amp: 1 thresh 0.5

//the spikegen tells the synapse when to activate based on the timetable addmsg from timetable to spikegen: type = INPUT, message = activation addmsg from the spikegen to the compartment’s AMPA element, type = SPIKE

// Next loop iteration or END

3. For all compartments receiving input…

Simulating in vivo-like synaptic input patterns in multicompartmental models

• What are in vivo-like synaptic input patterns?• When are such simulations useful?• How we do it using GENESIS• Some strategies for analyzing the results

– Matlab provides a flexible platform for customization and automation of data analysis.

– Movies can help you explain what’s going on in the model– Compare multiple models, each representing a distinct

alternative case.– Compare synaptic activity with output spiking for each

synapse. Look at synaptic efficacy as a function of location.

– Analyze model input-output relations

Movie: 20 Hz excitation, 2.5 Hz inhibition

Quantifying synaptic efficacy 1. Probabilistic method:

Efficacy = P (output spike | synaptic activation) / P (output spike)

Advantage: need only the output spike times and synapse timetables.

Disadvantage: a time window must be chosen (usually arbitrarily),and the best time window may vary with output spike rate.

2. Average synaptic conductance method:

Efficacy = peak of synapse’s spike-triggered average conductance

Advantage: no arbitrary time window needs to be selected

Disadvantage: must write the full conductance trace for every synapse during the simulation, then analyze each one individually.

Non-spiking dendritesSpiking dendrites,Uniform synapses

Nor

mal

ized

syn

aptic

eff

icac

yQuantifying synaptic efficacy

Nor

mal

ized

con

duct

ance

ave

rage

Non-spiking dendrites Spiking dendrites,Weighted synapses

Spiking dendrites,Uniform synapses

Nor

mal

ized

syn

aptic

eff

icac

y

Location-dependence of synaptic efficacy

Analyses of model spiking output1. Synaptic integration mode: interactions between excitation and inhibition

2. Variability of model spiking: synaptic –vs– intrinsic control of timing

Conclusions

• Many independent synapses can easily be added to a multicompartmental model using the synchan, timetable and spikegen objects in GENESIS.

• This method is useful for making inferences about how in vitro results will apply to the in vivo system and for studying single neuron input-output functions.

• Matlab provides a convenient platform for customizing and automating the analysis of the data.

Thanks to…

• Dieter Jaeger• Cengiz Gunay• Jesse Hanson• Chris Rowland• Lauren Job• Kelly Suter• Carson Roberts