Methodology for counting dorsal horn neurons

[cehanson]

Hey everyone, I work in a lab that usually studies free nerve endings and DRG, but my project is focused heavily on the dorsal horn. My PI has asked me to:
(1) find estimates for the number of neurons in each lamina of the dorsal horn
(2) find approximate diameters for the widths of each lamina.
Ideally, I should find separate estimates for both adult and juvenile mice.

I’ve been combing the literature for a week and the best insight I have is from Todd’s 2010 review, which states “there are approximately 400 projection neurons in lamina 1, ~5% of the total neurons in this lamina.” That suggests there are 8000 neurons in just the thinnest lamina!

I’ve since started wondering if it’s even possible/practical to count neurons with a handheld tally, like we always do with the DRG. The other relevant paper I’ve found (Haring, 2018), describes using a Matlab program to analyze the images of the DH.

All of that said, I think my questions are:
(a) If you work with dorsal horn neurons, how do you count cells mounted on slides for IHC or ISH?
(b) Is there a way to do it solely by hand/ImageJ, or is learning to use a fancy calculator like Matlab vital?
© How many neurons would you expect to see in each lamina?
(d) Do you have any good heuristics for approximating the widths of each lamina? My PI thinks I should be able to just use ImageJ’s measuring tools to estimate where each lamina begins and ends.
(e) Does anyone know a reference that might detail how this information changes during post-natal development?

Thanks for reading- since no one in my lab has much experience with the spinal cord, I’ve been so grateful to find this forum.

[achamess]

Sounds like a questions for @liz and @sshiers

But I’ll take a stab at some of this.

(a) If you work with dorsal horn neurons, how do you count cells mounted on slides for IHC or ISH?

You can do either. Depends on what methods you’re using to visualize your cells. If RNAscope, then it’s a mounted slide. You could also do IHC on a mounted slide if your antibody is good. The sections are thinner in mounted vs. free floating.

(b) Is there a way to do it solely by hand/ImageJ, or is learning to use a fancy calculator like Matlab vital?

Can you explain more what you mean? The first order of business in counting cells is deciding what is the the cell? In an automated fashion, this called segmentation. Scripts in any language, not just Matlab, will need a way to draw boundaries around whatever you are calling the cell. Most segmentation methods use nuclei as a proxy for the whole cell, because nuclei are very distinct with DAPI and can be segmented. When you do this by eye, you’re depending on your eye to determine the boundaries of the cell. Unlike DRG, DH has many more neurons and they’re small. If your population of interest is small, you can count by hand. if you’re look at every DH neuron, then automated makes sense. There are hundreds of cells.

Check out this new segmentation software:
https://www.cellpose.org/
I’m impressed and it works on DRG too.

You might also check out QuPath

QuPath

It has a segmentation module and is easy to count objects and cells. I really like it.

What will you be staining with? How will you mark all neurons? NeuN? Neurotrace? Something else?

How many neurons would you expect to see in each lamina?

Depends on segment. Would check the lit values.

(d) Do you have any good heuristics for approximating the widths of each lamina? My PI thinks I should be able to just use ImageJ’s measuring tools to estimate where each lamina begins and ends.

See this discussion.

How do people accurately and methodically draw laminar boundaries in spinal cord IHC images?

Many papers that do spinal IHC in the dorsal horn superimpose boundaries on the laminae. [27FF2] (http://anesthesiology.pubs.asahq.org/article.aspx?articleid=1944809) How do people consistently and accurately do this? What are the features or landmarks that you use? I know that certain molecular markers define laminar areas, for example PKC-gamma and lamina II inner and IB4 as lamina II outer. But in the absence of these markers, how do people do it? Neuronal density? that means you need to…

(e) Does anyone know a reference that might detail how this information changes during post-natal development?

Another great question for @liz

Also calling @tberta @gcorder @SamineniV

[james_maks]

I’m by no means an expert but I can also try to answer some of these questions based on my experience in the dorsal horn:

(a) If you work with dorsal horn neurons, how do you count cells mounted on slides for IHC or ISH?

IHC: I typically use NeuN staining or Neurotrace (Blue or FarRed work well) to identify neurons

ISH: If you are just interested in counting neurons in general, you might try Slc17a6/vGluT2 & Slc32a1/vGAT which should label all DH neurons (excitatory & inhibitory, respectively) and fill up the cell bodies nicely. Something similar to this figure from Hughes and Todd’s recent review. Unless you are able to find a high-expressing transcript that fills your cells of interest, you’re stuck using DAPI and attempting to extrapolate cell boundaries (or just using the DAPI mask).

(b) Is there a way to do it solely by hand/ImageJ, or is learning to use a fancy calculator like Matlab vital?

After you identify the neurons/nuclei, you can either manually count them, use ImageJ or other software to threshold/identify particles, or use a cell segmentation algorithm.

I second the cellpose suggestion by @achamess for automated cell segmentation. I’ve used it for both DRG and spinal cord and it works well to segment individual cells using NeuroTrace/NeuN. You could also apply this to the ISH experiment above, although this might take some optimization. There are a lot of other segmentation algorithms out there too; many of which have user-friendly GUIs which don’t require programming knowledge.

It is possible to count by hand (e.g. using the Multipoint Tool in ImageJ), although depending on how many sections you want to analyze this could be incredibly tedious. An alternative to manual counting is to threshold and analyze particles (example). This works in some situations but depends a lot on the quality of the staining.

The number referred to in the Todd 2010 review is 400 lamina I projections neurons (and ~8000 lamina I neurons) only on one side of the spinal cord so it’s really 16,000 lamina I neurons total but for the entire rat L4 segment. In a thin, transverse section you will likely have <100 lamina I neurons which is manageable by hand but if you’re analyzing other laminae across many sections it could be worth it to test out some more automated approaches.

(c) How many neurons would you expect to see in each lamina?

Depends on species and segment. Todd’s work is a good starting point.

(d) Do you have any good heuristics for approximating the widths of each lamina? My PI thinks I should be able to just use ImageJ’s measuring tools to estimate where each lamina begins and ends.

I like the discussion in @achamess 's response. If you can, staining with markers for the different laminae could help with this. I refer to this figure a lot:

https://dev.biologists.org/content/143/19/3434

[achamess]

Really fantastic input here @james_maks! Thank you so much. I learned some things too. Glad cellpose is working as advertised for you.

It might be nice to write up a protocol one day on this DH analysis. Maybe we could even host community protocols here eventually.

[achamess]

Replies from Twitter:

Olivia Davis (LivDavis117)