Today’s post is for those of you working with fluorescent reporters and trying to do intracellular staining. It’s hard, I know. I’ll offer my tips on what works, what doesn’t, and—as best I understand it—why.
First off, why do we lose GFP and other fluorescent proteins when we fix and permeabilize cells? In most cases, these proteins are being expressed independently or with an IRES, so they end up in the cytoplasm of the cell, free-floating, so to speak. They’re small proteins (GFP is 27kDa), only a bit bigger than cytokines, so if we go and punch big holes in the plasma membrane with detergents, they leak out of the cell. There’s a really nice paper on GFP and intracellular staining (link here), which shows that with a certain commonly used fix/perm kit, you get large amounts of protein released from the cell upon addition of the fix/perm reagent. I suspect this happens with all kits designed for nuclear staining.
The central conundrum: incompatibility of nuclear staining with cytosolic protein retention. Mouse splenocytes from animals transgenic for an iCas9-GFP reporter, gated on viable CD4+CD3+ T cells.
If on the other hand, you have GFP with a membrane tether, fused to another protein, or with a nuclear localization sequence (NLS) to get it beyond the harder to penetrate nuclear membrane, you’re much less likely to lose your GFP with fix and perm treatment. So, if you’re just getting started cloning your construct, consider either modifying your GFP localization or switching to a surface protein like CD2 or Thy1.1.
Now, if we understand why GFP is getting lost, we can work to change that. Adding the permeabilizing detergent with the fixative is going to open holes in the membrane while the fixatives tries to crosslink. If we first fix, then perm, we crosslink the GFP in the cytoplasm, preventing it from escaping when we subsequently permeabilize.
Formalin and triton separately versus together
Next, we can consider which fixatives to use and how to use them. Like cytokines, GFP is best preserved in the cell with higher concentrations of fixative for longer times (or higher temperatures). A 4% PFA solution (e.g., Thermo IC fix) for 30-45min at room temperature gives good preservation of GFP. The lower the concentration, the less GFP is retained. The less time used, the less GFP is retained (although anything above ~20min is good).
The catch is that 4% PFA is so good at crosslinking epitopes that many of your epitopes will be damaged and the nucleus will be hard to access, no matter how you permeabilize afterwards. And that’s the point of this paper. They switched to 2% formaldehyde (a methanol-stabilized solution) and got better results for transcription factors. And yes, there is truth to that. But, it doesn’t work as well or as generally as the paper suggests it does. It works well for T-bet, basically not at all for GATA-3, Bcl-6 or RORgT, and partially for Foxp3. What do I mean, partially? I mean when you use an independent reporter for Foxp3, such as surface-expressed Thy1.1 in Foxp3-Thy1.1 reporter mice, we see that the intranuclear staining for Foxp3 doesn’t actually stain all the Tregs, nor does the Foxp3 separate as well as we’d like. Overnight staining does improve the transcription factor staining, but doesn’t always fix the problem. Using smaller fluorophores can also help, and protein-tandem dyes tend to be problematic for Foxp3 more than other transcription factors. Cytokines, however, do work well with this method.
Completeness of Foxp3 protein detection as assessed by co-staining with the Foxp3-Thy1.1 surface reporter.
The formalin method is definitely an improvement for transcription factor staining compared to using PFA. Note that it doesn't work as well as dedicated transcription factor kits for most targets, though. T-bet is a notable exception, which consistently works much better with this approach.
Overnight staining helps a lot in this setting. Staining at 4C is probably your best bet as at higher temperatures you can get the appearance of some extraneous negatives, perhaps due to enzymatic protein degradation.
The retention of GFP and other cytosolic fluorescent proteins is also a bit worse with formalin than with PFA, although it can often be serviceable. In this case, we're looking at expression of a relatively weak GFP expressed from the ROSA26 locus in combination with Cas9 dependent on CD4Cre activity.
CD4+CD3+ viable T cells showing expression of a weak GFP reporter and varying levels of loss.
So, if you want to do a combination of transcription factors and GFP (or RFP, mCherry, YFP, etc.) you need to include some level of detergent in the fixative. This allows you to access the nucleus and reduces crosslinking. As a result, you get some loss of GFP, proportional to the concentration and strength of the detergent used to permeabilize.
Finding the Goldilocks zone
Kits designed to allow access to the nucleus will wipe out all your GFP, and it is generally gone, not recoverable. Caveat: there may be a combination of time, temperature and proportion of fixative to cells where you can preserve some GFP with the Foxp3 kit. I’ve had it happen, but I can’t reproduce it reliably. In any case, it’s poor at best.
That’s frustrating. What can we do about it? We probably need to accept an intermediate approach, hence finding the Goldilocks zone where both things work. One option that may get us a bit closer to this seems to be the BD Cytofix/Cytoperm reagents. If you look at the spec sheet for this kit, you'll see that it isn't meant to work for transcription factor staining. And, indeed, when used according to the instructions, it doesn't. If you leave it overnight, though, you can get this:
Not perfect, but getting better. Again, using smaller fluorophores (APC is 105 kDa, R-PE is 250 kDa) would help.
The BD Cytofix/Cytoperm kit produces similar results for GFP retention as 2% FA, and can produce comparable or better transcription factor staining when left overnight at 4C.
If we have a partial loss in GFP signal, we can amplify it using a fluorescently conjugated antibody. In conventional flow, AF488 conjugates are often used (link). The thing is, AF488 is spectrally distinct from GFP and can be separated pretty readily on spectral cytometers.
AF488 vs GFP unmixing with the anti-GFP-AF488 rescue. Mouse splenocytes. In this case, the signals are weak, so the separation is poor, but with stronger signals, unmixing this AF488-enhanced GFP will cause spillover errors in larger panels. The fixative used here is the BD Cytofix/Cytoperm kit.
AF488 vs. GFP spectra
So, if we do that, we’re likely to run into unmixing spillover errors due to using the wrong control or having a complex signature comprised of both GFP and AF488 that might be variable. There’s a better option. BB515 and Spark Blue 515 are both much more similar to GFP in their spectra. Not perfect matches, but quite close.
GFP, BB515 and Spark Blue 515 spectra traces (from Cytek Cloud)
BioLegend, who sells Spark Blue 515 conjugates, also produces anti-GFP clone FM264G, which gives very clean staining. I asked them to make us an anti-GFP conjugated to Spark Blue 515, and they did for a price not far off the normal catalog price. As a bonus, Spark Blue 515 is slightly brighter than AF488 and will have less spillover into other fluorophores off the 488nm laser.
AF488 vs. Spark Blue 515 anti-GFP enhancement of iCas9-GFP (expression dependent on CD4Cre, so most T cells) in mouse splenocytes treated with the BD Cytofix/Cytoperm kit and stained overnight.
The Spark Blue anti-GFP isn’t available yet in the catalog. If you want to buy it, contact your local BioLegend rep. The more people who ask, the quicker it’ll get added to the catalog.
What about other fluorescent proteins? These principles apply to all of them. If you want to amplify signals from others, you need to understand what the protein actually is. FPbase is a good place to start. For instance, YFP and Citrine are both modified GFP, so an anti-GFP would recognize these. To amplify YFP, you’d need to conjugate the anti-GFP to a fluorophore with a similar output to YFP (e.g., iFluor514). mCherry and many “RFPs” are derived from dsRed. The Rockland anti-RFP works pretty well for flow, albeit with some background that needs to be blocked with rabbit serum. You’d then need to find a fluorophore closely matching mCherry (perhaps AF594) or TdTomato (perhaps CF570).
Relevant papers:
Black-necked Stilt, Costa Rica
thank you for this post. so well executed. have you considered a non-permeabilizing fixation first to trap the GFP and then do a usual fixation procedure with the eBioscience FoxP3 kit to get kind of the best of both sides? or do you think it might be rather detrimental than helpful?
Excellent work, as usual.