Interpreting the results of electrotonic analysis

[bril27]

I have tried to do a neuromorphic rendering of the electrotonic transform of a mitral cell model using the Impedance\Shape tool. For a 10-Hz input with the soma as the reference, I find the following Vin and Vout results. (Please note that the orientation of the neuron in the bottom plot is a little off to an angle compared to the top (original) plot.)

Vin:
https://drive.google.com/file/d/1oJi0cs ... sp=sharing

Vout:
https://drive.google.com/file/d/1SFynUo ... sp=sharing

For Vin, is it correct to interpret this as a widespread attenuation of the input signal from all of the dendrites (apical and basal)?
For Vout, I am not clear as to why there are disconnected portions of dendrites - how should this be interpreted? On the other hand, for those segments that do not appear in this neuromorphic rendering, should we assume that those are electrically connected to the soma with little to no attenuation?

Thank you for your help!

[ted]

Two suggestions. First, regarding how to interpret neuromorphic representations of electrotonus: work through this exercise https://nrn.readthedocs.io/en/latest/co ... c-analysis

<< note added after my initial post: at present, the "Shape" link in the list of tools for electrotonic analysis points to documentation of the Shape class. That will be fixed "real soon." For now, to see documentation of the electrotonic analysis "Shape" tool just scroll down the page to find the heading "The Shape Tool" >>

and maybe also read
Carnevale et al.. The electrotonic transformation: a tool for relating neuronal form to function. In: Advances in Neural Information Processing Systems, vol. 7, edited by G. Tesauro, D.S. Touretzky, and T.K. Leen. Cambridge, MA: MIT Press, 1995, p. 69-76. Preprint https://proceedings.neurips.cc/paper/19 ... -Paper.pdf

Second, regarding your question "why there are disconnected portions of dendrites" in the Vout rendering of the model cell. The first thing to do is discover whether or not all dendritic trees are electrically connected to the soma. This is easy to do: create a graph that renders all neurites in colors that correspond to membrane potential, inject a large and prolonged depolarizing current into the soma, and watch what happens. Trivial with NEURON's GUI--requires no code writing at all. Anybody who's allergic to the GUI can probably do this fairly quickly with hoc or Python.

Or execute topology() (h.topology() for all you Pythoninstas) and examine the resulting output. If there is only one root section, and that section is called soma, then all dendritic trees are electrically connected to the soma.

"OK, suppose they're all electrically connected to the soma. Why does the Vout transform look like they aren't?"

That's because electrical connectivity is specified by "connect" statements, but position in space is controlled by pt3d statements. AND there is something peculiar about the code that defines the geometry of your particular model cell. I'd be glad to look at it and tell you what that "something peculiar" is, if you like.

[ted]

Neuromorphic renderings of the electrotonic transformation might be pretty and seemingly intuitive, but many users find logA vs. X plots to be far more informative.

[bril27]

Right, it seems that there are disconnected pieces in the model itself, which I downloaded here:
https://neuromorpho.org/neuron_info.jsp?neuron_name=2M3

Here's a snapshot of after injecting a long-duration 10 nA current into the soma:
https://drive.google.com/file/d/1q5O5hX ... sp=sharing

And this is the topology:
https://drive.google.com/file/d/1fOk2ra ... sp=sharing

Please advise on how I could possibly connect all of the broken segments.

[ted]

From the printout generated by topology() that you provided, it looks like
1. there is only one root section
2. the root section is called soma
3. there is one axon and six dendritic trees
and
4. the proximal ends of the axon and all dendritic trees are attached to the 1 end of the soma

In the morning, I'll get the source morphology from NeuroMorpho.org and see what else I can discover.

[ted]

Downloaded a zip file from NeuroMorpho that was supposed to contain a couple of "log files" and two morphology files: one that NeuroMorpho calls "Original" and another called "Standardized". Expanding the zip file revealed two morphology files, one called 2M3.asc (in NeuroLucida text i.e. "ASCII" format, presumably the original morphology data) and the other called 2M3.CNG.swc (presumably NeuroMorpho's "standardized" morphology data).

Next I ran nrngui and used Import3d to import 2M3.asc
This generated the following output from NEURON's hoc interpreter:

Code: Select all

5290 lines read
./2M3.asc problems
Main branch starting at line 176 is outside the soma bounding boxes
  Making a logical connection to center of nearest soma
Main branch starting at line 2522 is outside the soma bounding boxes
  Making a logical connection to center of nearest soma
Main branch starting at line 2707 is outside the soma bounding boxes
  Making a logical connection to center of nearest soma
Main branch starting at line 3541 is outside the soma bounding boxes
  Making a logical connection to center of nearest soma

Now, that sounds like a fairly clear statement of what to expect when viewing the resulting model cell in NEURON: the proximal ends of four branched structures (dendritic or axonal trees) will appear to be physically separated from the soma.

So there you are. The fault lies in the original morphometric data. Blame a lack of standardization across laboratories about how to deal with somas when collecting morphometric data from neurons (this has been a striking and chronic problem for decades).

I should mention that, in addition to printing the above messages, NEURON popped up a panel that displayed the message
2M3.asc: File translation problems. See the messages on the terminal
and offered a Continue button. Clicking on that button brought up a panel with a drag bar labeled "Possible root branch errors." This panel reported
"Default logical connection to nearest soma" (which in this case should probably be accepted)
but it also offered a way to override this default action: four checkboxes that could be used to connect each of the proximal ends to the 1 end of the "closest parent in the xy plane". I suspect that would not be the correct choice for this particular cellular morphology.

Yet another alternative would be to accept the "standardized" morphology by using the Impor3d tool to import 2M3.CNG.swc (this file is in the "CNG version" subdirectory)
Do that, and you'll see no warnings or panels--but you'll still get a model cell that looks peculiar because the proximal ends of the dendritic and axonal trees don't appear to originate from the soma. Again, that's the fault of the original morphometric data.

Final question: are the dendritic and axonal trees actually electrically connected to the soma? I set Ra to 100 ohm cm, cm to 1 uf/cm2, inserted no other membrane properties, and discretized the cell using the d_lambda rule (with d_lambda = 0.1), applied a 1 nA current step at the soma that started at t = 1 ms, and examined a false color Shape plot using NEURON's default color scale--and found that the entire cell became quite yellow within 150-200 ms, which indicates that all sections are indeed electrically connected to the soma.

Rotating the cell reveals a small amount of backlash artifact (slop) along the z axis--far from the worst I have seen. Checking diam3d values shows that the smallest and largest measurements are plausible (but I always wonder about the accuracy of values < 0.3 um).

[bril27]

Thanks so much, Ted. I think I will go the .asc route in the nrngui and run the Electrotonic Analysis tutorials. Great learning experience with the use of morphometric data.