Page 1 of 5

pyramidal spine-neck EPSPs

Posted: Mon Nov 28, 2022 10:30 am
by davidhubbardmd
I am using the GUI to build a model of the V1 pyramidal neuron spine neck's prolonged EPSP. I am looking for a consultant or tutor to assist me. I am a neurologist doing fMRI research.
David Hubbard MD
San Diego

Re: pyramidal spine-neck EPSPs

Posted: Fri Dec 02, 2022 10:05 am
by davidhubbardmd
Spine-neck Neuron model draft
3 segments: spine-neck, spine-head, dendrite.
Biophysics: no HH, best available ion and glutamate channel data
Point processors: Alpha Synapse in spine-head and IClamp on spine-neck. Fields?
Graph: waveform of voltage and ion currents along the 1-2um length, 0.5um diameter spine-neck segment.
Testing questions
#1 response to synaptic activity
relation of synaptic activity to spine-neck EPSP waveform including threshold and duration.
#2 response to electric field stimulation
Threshold to increase pos. ion influx in spine-neck
Intracortical and TCS parameters predicted to stimulate phosphene-like visual sensations
for intracortical (<500um) and transcranial electrodes 10sec, 50ms duty cycle
Goal:
#1 Stimulation parameters for intracortical and transcranial devices.
#2 Stimulation pattern to display, e.g., the numeral ‘7’, in phosphenes, using a network of +ion-influx into spine-necks, distributed retinotopically.

Re: pyramidal spine-neck EPSPs

Posted: Mon Dec 05, 2022 12:12 pm
by ted
Not to open a different discussion, but in the experiments by Cornejo et al. 2022, why were "spine-only events" observed only as spontaneous events? What accounts for the failure to elicit them by photostimulation? Maybe I missed this, but did they find a subset of spines in which both spine-only and not-spine-only events were observed?

Much of what is outlined in your draft could be accomplished with NEURON's GUI tools such as the CellBuilder.

I'd suggest starting with a dendrite ~ 100-200 um long (approximate length of unbranched neurites in the basilar tree of cortical pyramidal cells). Maybe longer, to prevent boundary conditions (termination or branching at the distal end, termination or attachment to the parent neurite or soma at the proximal end) from affecting membrane potential at the site of spine attachment. The spatial discretization parameters of the dendrite and spine neck could be set according to the d_lambda rule, which is a built-in option of the CellBuilder.

Wouldn't include fields at this stage. Your first goal should be to explore and develop insight into how membrane potential evolves over space and time following synaptic activation.

Re: pyramidal spine-neck EPSPs

Posted: Mon Dec 05, 2022 12:43 pm
by davidhubbardmd
Excellent, ty.

Re: pyramidal spine-neck EPSPs

Posted: Mon Dec 05, 2022 1:57 pm
by davidhubbardmd
Here are the beautiful videos from the Cornejo et all '22 supplement.
https://youtu.be/oxxBzHaXjLg
https://youtu.be/3vvuQg2tcFI

Re: pyramidal spine-neck EPSPs

Posted: Mon Dec 05, 2022 2:46 pm
by davidhubbardmd
The photostimulation of the spine head decayed into the dendrite, consistent with synaptic activation. The photostimulation of the dendrite entered the spine head with no attenuation, consistent with their “compartmentalization” conclusion. During whisker stimulation, action potentials were observed but also subthreshold activity of segments of dendrite and associated spines. Spine-only depolarization was not changed during sensory stimulation but did occur “independently in the absence of dendritic or somatic activity.”

Re: pyramidal spine-neck EPSPs

Posted: Tue Dec 06, 2022 4:03 pm
by davidhubbardmd
I ran the following in the GUI:
soma: L 30, Diam 30, hh
basilar_dendrite: L 300, Diam 1, d_lambda
spine_neck: L 100, Diam 1, d_lambda (attached to end of basilar_dendrite (unable to reposition away from distal end)
spine_head: L 30, Diam 30
Alpha Synapse(0) at spine_head(1), Onset 0.5, tau 1, gmax 0.05
RunControl: Init -65, Cont til 5, Cont for 1, t(ms) 0, Tstop 5, dt 0.025, Points 40

Graph: got a nice wave that peaked at -20mV at 1 ms and decayed.
Tried in vain to post the waveform here :)
Next: increase Alpha Synapse to simulate a repetitive volley from the thalamus for 50 ms.

Re: pyramidal spine-neck EPSPs

Posted: Wed Dec 07, 2022 5:06 pm
by davidhubbardmd
Alpha Synapse seems to fire just once.
IntFire2 seems to allow repeat firing, so I'm trying to send enough depolarizations from the spine_head that my voltage graph at spine_neck(0.5) will detect it.

Re: pyramidal spine-neck EPSPs

Posted: Thu Dec 08, 2022 5:32 pm
by davidhubbardmd
IntFire1, using GUI
Topology Geometry: soma L= 30 Diam = 30
Graph: v(0.5) at soma(.5)
IntFire1[0] at soma(.5), tau(ms) = 10, refrac(ms) = 5, refrac = 5, m = 0
RunControl: “RunControl: Init -65, Cont til 5, Cont for 20, t(ms) 0, Tstop 20, dt 0.025, Points 40
I got no response.

Re: pyramidal spine-neck EPSPs

Posted: Fri Dec 09, 2022 12:48 am
by ted
First, a side comment regarding this:
spine_neck . . . attached to end of basilar_dendrite (unable to reposition away from distal end)
True, the CellBuilder only allows the 0 end of a section to be connected to the 0 or 1 end of another section. So how does one use the CellBuilder to implement a model in which a spine is attached somewhere along the length of a parent neurite? Represent the parent by two sections connected in series, each section with appropriate length, and attach the child to their junction.* You'll see this in the CellBuilder that is part of the demo model I'll email to you.

*--This is also a useful strategy for implementing models in code if you want the location of the spine to be unaffected by changes of nseg.

End of the side comment.
Alpha Synapse seems to fire just once.
AlphaSynapse generates a single "synapse-like" conductance change that happens at a fixed, user-specified time. The practical way to implement repetitive synaptic activation is with an event-driven synaptic mechanism. ExpSyn with e = 0 mV and tau = 3 to 5 ms is a decent first-order approximation to an AMPAergic synapse.

But you'll also need two more things: an instance of the NetStim class to generate events, and a NetCon to convey events from the NetStim to the ExpSyn.

And at this point, the complexity of your model requires exploiting the strengths of the GUI and code. The GUI is helpful for creating and managing the properties of the NetStim and the ExpSyn (and to reposition the ExpSyn), but a bit of code is required to make the NetStim drive the ExpSyn.

To illustrate this, I used the CellBuilder to implement a model cell similar to what you described in a previous post. In the CellBuilder note:
0. section orientations were deliberately chosen to make it easy to identify where connections are located.
1. "head" appears to be a keyword, so the spine head needs to be called something else. I chose kopf because it's short.
2. subsets are used to simplify specification of the model cell's properties.
3. Ra is set to 100 ohm cm, which is in the range of values used for cytoplasmic resistivity.
4. g_pas is set to 1e-4 S/cm2, for a membrane time constant of 10 ms (assuming cm = 1 uF/cm2)
5. e_pas is set to -65 mV because the soma uses hh, which wants to rest at -65 mV. This makes initialization easy: just initialize membrane potential in all sections to -65 mV. Using a different value for e_pas would produce a model that has nonuniform resting potential and requires custom initialization.

I used the Print and File Window Manager (PFWM) to save the CellBuilder (in Continuous Create mode) to a session file (cell.ses) all by itself.

Then I created a RunControl panel, a "voltage axis graph" that shows v at the middle of the soma, and a "space plot" that shows membrane potential as a function of distance along two paths in the cell: one (red) from the distal end of the basilar dendrite to the 1 end of the soma, and the other (blue) from the distal end of the spine head to the 1 end of the soma. At rest the blue line lies right on top of the red line. I saved these three items to a ses file (basicrig.ses) by themselves.

Next, I used PointProcessManagers to create one instance each of the NetStim and ExpSyn classes. The ExpSyn I moved to the 1 end of kopf; it doesn't matter where the NetStim is located. I set the ExpSyn's tau to 3 ms, and saved these two PointProcessManagers to a file called syndrive.ses.

Now we get to the part where the GUI is combined with a bit of code.

With a plain text (programmer's) editor I created a file called init.hoc that contaned these statements:
load_file("nrngui.hoc")
load_file("cell.ses")
load_file("basicrig.ses") // RunControl, v vs. t, and v along paths
load_file("syndrive.ses") // PPMs as NetStim and ExpSyn

Then at the system (OS) prompt I executed the statement

nrngui init.hoc

and saw that NEURON recreated the model and all of the GUI tools that I had used. Running a simulation showed that, as usual, soma.v wobbled a little bit during the first 15 ms or so before settling to -65 mV. Clearly the first synaptic activation should be delayed until t ~ 15 ms.

But how to get the NetStim to send events to the ExpSyn? At NEURON's oc> prompt I entered the following statements, one at a time:

objref ns, syn, nc
ns = NetStim[0]
syn = ExpSyn[0]
nc = new NetCon(ns, syn)
nc.weight = 1e-3

The first statement creates variables with convenient names that can be used to refer to the class instances. The second and third statements make ns and syn refer to the NetStim and ExpSyn. The fourth statement creates an instance of the NetCon class that will convey events generated by ns to syn.

I ran a simulation, saw nothing, and noticed that Tstop was 5 ms but the NetStim's first event doesn't happen until 50 ms. So I changed Tstop to 25 ms and the NetStim's start parameter to 15 ms, and ran another simulation. Still nothing. But of course--the NetCon's default weight is 0, so the events delivered to the ExpSyn don't affect synaptic conductance.

Executing this statement
nc.weight = 0.001
made a single activation generate a synaptic conductance transient with maximum amplitude 0.001 uS = 1 nS. That's strong enough to elicit a ~0.4 mV EPSP at the soma.

This was a reasonable starting point. I used the text editor to add the following statements to the end of init.hoc:

objref ns, syn, nc
ns = NetStim[0]
syn = ExpSyn[0]
nc = new NetCon(ns, syn)
nc.weight = 1e-3 // just for a test

And in the RunControl panel I changed Tstop to 200 ms, then added kopf.v(0.5) to the v vs. t graph, and saved the RunControl panel, the v vs. t graph, and the v vs distance graph to basicrig.ses.

I'll zip up these files and email them to you so you can see and try the result.

Re: pyramidal spine-neck EPSPs

Posted: Fri Dec 09, 2022 1:21 am
by ted
IntFire1
. . .
I got no response.
That is as it should be. The IntFire* classes built into NEURON are artificial spiking cells. Their class names appear in the PointProcessManager's list of point process classes for historical reasons: when they were first added to NEURON, it was convenient to do that in a way that took advantage of existing code that was originally used to implement IClamps, AlphaSynapses etc.. However, unlike those classes, artificial spiking cells do not generate or respond to electrical signals, and the only way they can interact with other artificial spiking cells or biophysical model cells (model cells that are built from sections) is by sending or receiving events.

IntFire2 can generate repetitive events, but the NetStim class provides direct control over number of events, start of the event train, inter-event interval, and randomness.

Re: pyramidal spine-neck EPSPs

Posted: Sat Dec 10, 2022 7:11 pm
by davidhubbardmd
Thank you Ted, adding the event control to the Alpha Synapse worked nicely. I got the small depolarizations of 4mV at the spine_head (kopf) and 0.4mV at the soma.
Next, at the oc> prompt, I increased nc.weight from the initial 1e-3 (0.001) until I got an AP at 1e-1 (0.1) which peaked at +40 mV. The spine_head (kopf) depolarized from -65 to -28mV, without an AP of course.
Next, leaving the weight at 1e-1 (0.1), I added the spine_neck to the graph and ran the experiment again. The AP increased the spine_neck voltage to about -8mV. After the AP, the spine_neck voltage and spine_head (kopf) oscillated together for 100 ms, with the spine_head about 10 mV less depolarized at about -40mV.
Next, I raised the weight to 1e2 (100) [not 10 my error] and got a roughly smooth plateau at about -10mV for the spine_neck. Further weight increases had no effect.
Are weights determined by the density of AMPA channels and their degree of opening? I'd like to see what the ion currents look like.

Re: pyramidal spine-neck EPSPs

Posted: Sun Dec 11, 2022 10:53 am
by davidhubbardmd
here's an image of the last experiment.
https://drive.google.com/file/d/1L7Y5zj ... sp=sharing
I note the repetitive activation of the Alpha Synapse fires only one AP at the soma, but maintains the voltage potential at 0 mV for the spine_head ('kopf') and at -15mV for the spine_neck. To trigger the soma, a more realistic setup might be several spines rather than a single heavily weighted one, but my interest is in spine_neck EPSPs, not the highly over-rated axon :) .

Re: pyramidal spine-neck EPSPs

Posted: Wed Dec 21, 2022 4:12 pm
by ted
increased nc.weight from the initial 1e-3 (0.001) until I got an AP at 1e-1 (0.1)
A single activation of a single ampaergic synapse onto a mammalian pyramidal neuron elicits a conductance transient that has a peak amplitude of about 1 nS. A single event delivered to an instance of NEURON's ExpSyn mechanism will produce a conductance transient that has a peak amplitude (in units of uS) that is numerically equal to the event's weight. This means that
* an event with weight 1 will elicit a conductance transient with peak amplitude 1 uS
* if the weight is 0.001, the peak conductance will be 0.001 uS = 1 nS
* if the weight is 0.1, the peak conductance will be 100 nS, which is about two orders of magnitude larger than what has been observed in conventional experiments (e.g. single electrode voltage clamp).

A very recent paper by Moradi et al. (available at no charge from https://www.nature.com/articles/s42003-022-03329-5), which used a novel application of deep learning to estimate unitary synaptic conductances, reports peak conductances in the range of 1 to 3 nS (see their figure 7). Achieving a peak conductance change of 100 nS at the head of a spine would require that at least 30 separate presynaptic terminals from excitatory neurons be attached to that spine, that all of them be activated simultaneously, and that the spine be so richly endowed with glutamate receptors (coupled to channels) that they aren't completely saturated after one or two terminals release transmitter. I am not aware of any experimental reports that individual spines are driven by that many presynaptic glutamatergic terminals (one terminal per spine seems to be the rule, although there is evidence that some spines are also innervated by a gabaergic terminal).

WRT spine anatomy, spine necks generally tend to be < 5 um long, although longer necks have been reported in autopsy or biopsy material from human brain in the setting of severe developmental disorders.

Something that hasn't been much studied is the effect of cytoplasmic organelles on electrical properties of the spine neck. Endoplasmic reticulum sometimes extends from the dendritic shaft into spine heads; does it occupy a significant fraction of the cross sectional area of the spine neck? if yes, does its membrane interfere with current flow along the neck? What about mitochondria or other organelles?

Re: pyramidal spine-neck EPSPs

Posted: Wed Dec 28, 2022 1:28 pm
by davidhubbardmd
Ty Ted, I returned the AMPA synaptic weight to 0.001 for the realistic conductance of about 1nS.

Then I changed the spine anatomy to reflect accurate measurements:
spine_head (kopf) 0.7um,
spine_neck length to 1,
diameter to 0.3;

I drastically increased the neck resistance from 100 ohm to 100 Mohm (1e-9) based on the published range of about 25-800 Mohm).

I added the spine_neck voltage to the graph, and reduced the Net Stim number of synaptic activations from 10 to 3.

I’ve linked the plot here:
https://drive.google.com/file/d/1hEc2VS ... sp=sharing

The spine_head (kopf) and spine_neck had identical EPSP waveforms peaking at - 57 mV. I was not expecting that given that the resistance in the neck is much greater.

Also surprising to me, the soma had a small response of about 2 mV.