RNAscope (In situ hybridization) to study gene expression in pain-related tissues


In situ hybridization is a powerful method to interrogate gene expression in tissues. It is known as being a challenging procedure due to the many steps where errors and obstacles arise. We’ve recently begun to use a commercial system called RNAscope from ACD Bio that gets around many of the pitfalls of traditional ISH.

I’ve used it a lot now in the spinal cord and DRG. Below is an image of the gene VGAT (Slc32a1) in the dorsal spinal cord. There is nice expression, and most importantly, it was a quick 5 hour protocol for results. I purchased the probe, performed the procedure (with some modifications) and then got the data I needed.

Downsides are that the reagents are expensive (although the time savings make it worth it), and that there still needs to be some optimization with respect to tissue preparation. It is this last step that I think is worth discussing here.

In our lab, we’ve settled on using the Fresh Frozen Tissue Preparation procedure, even though we use fixed frozen tissue. We perfuse our mice with fresh 4% PFA, then post-fix for 2 hours. 30% sucrose overnight (RNase-free) and then slice at 14 um onto superfrost plus slides. Then we use take the sections forward using the Fresh Frozen protocol, as opposed to the one from ACD that is indicated for Fixed Frozen. In addition, we skip the initial 15 min post-fix in the Fresh Frozen protocol. Everything else after that stays the same.

If you’ve used RNAscope, please share your experiences and whatever tips/tricks you’ve picked up.

Best antibodies to label different neuronal populations
Iba-1 and GFAP antibodies

Our lab has also used RNAscope, but on dissociated mouse DRG neurons that were fixed with 4% PFA /4% sucrose and stored for up to 1 year at 4 degrees. As was mentioned, the assay is very straightforward, and I’ve gotten really nice, clean signal for my target probes. However, I am struggling with how to quantify target coexpression.

RNAscope offers a semi-quantitative scoring guideline (https://acdbio.com/technical-support/solutions) with a scale of 0-4 based on number of RNA copies per cell and recommends adjusting the scoring criteria based on the gene expression level. Has anyone used this semi-quantitative scale to evaluate target coexpression? Or have you worked out your own criteria? I would be interested to hear people’s thoughts/approaches on this, especially with respect to establishing criteria for defining a cell as positive for transcripts that may differ in expression levels (i.e. 0-15 RNA copies vs. 0-40 RNA copies).

RNAscope also offers quantification software, but we weren’t planning to invest in the software at this point in the project.

Thanks in advance!


Hi @TSheahan. Thanks so much for joining the forum and for your question.

Quantification is definitely a tricky subject, one that I don’t have a perfect answer for. But the answer will depend in part on what you’re trying to quantify. If you’re looking for a binary answer (express/not expressed) then it’s a matter of setting a threshold for your signal on your image, and then counting the positive cells. Now, how to set the threshold? That’s partly art. The more stringent you set it, you may have false negatives, but you will gave more true positives. I err on the side of more stringent so I can be sure I’m really seeing true positives. But then of course I’m going to miss real ones too. It’s a tradeoff. Since you’re dealing with dissociated cells, you should see very little background outside the cells. You can afford to be less stringent perhaps. But you should run the RNAscope negative control probe. This will give you an objective measure of what amount of signal is just nonspecific binding of the non-probe reagents. You could then subtract this level of signal as “background”.

RNAscope is quite sensitive, and so you can see a single puncta in a cell or sometimes outside a cell. I noticed with the original multiplex fluorscent kit (v1) there was quite a lot of stray dots outside of cells and it made quantification hard. The new v2 kit, which uses TSA amplification, has very little background and will help things a lot.

Now, if you’re looking for quantitative results (not binary, but degrees of expression), then you need to count puncta. For low expressed genes, this can work because you see single or a few copies. But for highly expressed genes (like my example above), there is so much mRNA that you just see clusters. Then quantification is not possible.

For an example of the puncta counting approach in neurons, see this paper from Henry et al: https://elifesciences.org/articles/09800 They also recommend some image-J software.

If you’re doing multiple colors, and want to see multiple colocalizations (binary), I don’t have a great answer. See this discussion: Colocalization analysis of cells - What’s the best way? (ImageJ, Photoshop?)

But @mny3 wrote a sweet MatLab script that kind of does it. he may be willing to share.

I hope that helps. Please share your experiences with us and whatever you learn.


Another response on Twitter


Thanks @achamess! Just enrolled in a very timely webinar with RNAscope on data analysis and representation. I’m interested to hear their recommendations as well.



I’ve recently completed some successful RNAscope experiments (with triple label IHC) on fresh frozen lumbar DRGs and spinal cord. Tissues were collected, frozen on dry ice (in OCT), sectioned immediately at 20um on a cryostat (onto the slide), dehydrated in EtOH, and fixed in 10% formalin for 15 minutes.
The tricky part I have found is with the protease treatment. If you treat your sections for too long, you increase background / get “holey” tissue / etc (details are highlighted on the ACD website). I found 5-7 minutes with my tissue prep works the best (7 minutes is even pushing it). Likewise, with this shortened period of fixation and protease treatment, I was able to preserve antigen epitope and conduct triple label immunohistochemistry on top of my ish for a cell-type specific analysis.

I measured cell diameters and looked at cell populations expressing my target mRNA. (That way I could avoid quantifying puncta since as it was pointed out, there can be multiple clumped together in one cell).


Hi @sshiers
Thanks for the post and for sharing your tips. You’re absolutely right about the protease time and fixation time being critical variables. You used fresh frozen, which they have a protocol for, although it seems you’ve reduced the protease time substantially.

I actually apply the fresh frozen protocol to fixed frozen, and this seems to get good results. I do a transcardial perfusion, 2 h (no more!) post-fix in PFA, sucrose, and then cut at 14 um. Then I proceed with the fresh frozen pretreat without the 15 min fixation step (since my tissue is already fixed). I do 30 mins of Protease IV. This gets good results without damaging the tissue. I try to avoid the whole boiling antigen retrieval business.

The new v2 kit is amazing. With v1, I always had some spotty signal outside the tissue and higher background, and low expressed genes were hard to see in DRG or spinal cord. The v2 kit really improves that.


Mouse or rat?

Yes, I shortened the protease time due to my short fixation period as I didn’t want to disrupt my IHC label. I was trying to find that sweet spot between too much fix and too much protease. :smile:
Have you tried a shorter perfusion with shorter protease time? Could cut down on your experiment time - a 2 hour perfusion seems really long to me.

My mouse perfusions for when I do IHC are 8 minutes long with a 1 hour post fix (brain IHC). No AR required and staining looks stunning.


Hi @sshiers
That’s 2 hour post-fixation (piece of cord in PFA in a tube @ 4C), not perfusion. That’d be an insanely long time and I’d be in grad school forever :slight_smile: I perfuse for 3-5 minutes with PFA, take out the cord or DRG, the do the post-fix.


Gah read your message wrong. Phew :slight_smile:


@esypek Hey, have you used the v2 kit with three colors, including Fluorescein?

Cy3 and C5 look awesome, but Fluorescein is garbage. The issue is definitely the fluor, since we can use the same probe, same tissue, and get wildly different results.

Maybe our fluorescein is bad? Green is never my favorite color for anything.


An update: At the suggestion of @ACDbio support, we tried another green TSA fluor, Opal 520. Now we can see much better signal. It’s still not as awesome as Cy3 and Cy5, but it’s visible. I would save green fluors for the highest expressing genes. I think also we will increase the concentration of Opal 520 to 1:750 vs. the recommended initial dilution of 1:1500.

@esypek @tberta