FACS spinal cord microglia


Hi all,

Wondering if anyone has had experience isolating microglia from spinal cord tissue? I’ve tried various protocols (Dounce homogenization and enzymatic digestion) with relatively low yield. Looking to do RNA seq afterwards and am having too low of a yield to do so. I gate with CD45 mid/CD11b+/CX3CR1. Any help on this is greatly appreciated! Of course these protocols works beautifully in brain, way less successful in spinal cord.



I’ll preface this by saying I’m not a microglia expert, but I work with spinal cord dissociation quite a bit. Hopefully some microglia experts like @esypek can chime in.

Sounds like you’ve been following the standard protocols. Spinal cord is definitely harder in most ways compared to brain, largely, IMO, because the large amount of axon and myelin. Dissociating neurons from spinal cord is very difficult too. This is why I use nuclei for nearly all my genomics work where I need to get some cellular RNA out. Although there are reasons to not do that as well.

I’m sure you’ve seen this paper, but in case you haven’t:

And here, Haring et al. show an optimized protocol for spinal cord dissociation with the targeting of neuronal cells, but likely will help for microglia:

Marques et al. isolated oligodendrocytes from spinal cord for single cell:

Some questions:

  1. How many microglia do you need?

You say

Looking to do RNA seq afterwards and am having too low of a yield to do so

What yield do you think you need? The RNA-seq library prep will determine this. And in my mind, too low is no longer an issue. If we can do RNA-seq from a single cell, we can do anything. I routinely do sub-nanogram levels of RNA using the Smart-seq2 protocol or the Nugen SoLo (https://www.nugen.com/products/ovation-solo-rna-seq-system). If you want to do amplification-free, or you’re trying to pull down ribosomal mRNA, that’s a different story. You need to increase your yield, which means more cells. But if you’re just trying to do regular gene expression profiling, the low-input methods I mention here will take whatever you give them (down to a single cell).

Now, if you want to try to increase cell yield, there are a number of ways:

  • Pool animals
  • Optimize your dissociation protocol to (1) increase viability (2) increase number of cells liberated from the solid tissue, or ideally both. In your protocol you’re using, it could be that you’re freeing up a good number of microglia but they’re not surviving (and thus, viability is the issue). In that case, you want to do things that help with that (oxygenating your buffers, working fast, using the gentlest protease, douncing more gently). Do you check your preps at each step to get a sense of where our losses are? What number of viable cells do you get?

These questions lead me to the next question/recommendation. Are you willing or able to use a transgenic line such as Cx3cr1-GFP or Sall1-GFP? That would allow you to remove the antibody labeling steps, which will likely increase viability (reduce time to FACS) and make your life easier. Also, because they’re fluorescent from the get-go, you can optimize your protocol better because you can look under the scope at each manipulation you do. The FACS sorting will likely be cleaner too since fluorescent proteins don’t have the background binding issues that antibodies do. If you’re not against using a transgenic, I recommend this. Note, however, that some researchers don’t like the transgenic lines because they believe that the physiology of the microglia are affected by the transgenes. This is particularly true of Cx3cr1-GFP, which is a knockin. So effectively you’re working with a heterozygote. Alternatives are a Cx3cr1-GFP BAC. Note sure if there are others.

Here is the Sall1 mouse:

Lastly, if you’re not committed to using cells, you can use nuclei, in which case you’ll still need a transgenic to label the nuclei. Benefits of nuclei are that the dissociation is easier and likely less biased in terms of cell type capture since you’re not having to keep cell membranes intact. You just dounce and go. You can use frozen tissue too. I use that strategy here to get SST+ neuronal nuclei from the DH:

One additional consideration. If you’re going to isolate live microglia and look at their transcriptomes, your dissociation is going to induce some gene expression just from the prep. This is an important issue for everyone doing single-cell or bulk RNA-seq on dissociated cells. We cannot ignore it.

Discussed here:

If your question for RNA-seq revolves around seeing genes induced by injury or inflammation, you’d do best to reduce this dissociation-related gene induction as best you can. The method used here in Hvratin 2018 would be advisable (this goes for everyone doing cell dissociations, IMO):

So summary:

  • Decide what “too low” is for RNA-seq. If this is your main concern, it might not even really be a barrier depending on the library prep you use
  • Consider using transgenic microglia lines
  • Systematically optimize your protocol based on the published ones, paying attention to artefactual gene expression
  • Consider nuclei if you’re not committed to cells

@tberta @liz


Hi everyone, just wanted to follow up on this. I’ve done quite a bit of tweaking of various protocols the last few weeks, and I have a protocol that works beautifully. I am getting around 50,000 microglial cells with a pooling of 4 lumbar spinal cords. More than happy to share if anyone needs it!

The biggest thing is bubbling your solutions and keeping everything cold. I use a combo of mechanical and enzymatic digestion, unfortunately can’t seem to avoid enzymatic digestion which I know can be problematic. Mechanical alone does not do the trick.


That’s great @ShanTan! Thanks for coming back to share the update and for being willing to share your protocol. Good luck with your ongoing experiments!