Cell type-independent cell surface staining on FFPE to generate secondary cell masks

Hi there,

This is my first contribution to the Microforum and it deals with the following problem:
I am planning to perform multiplexed immunofluorescence on FFPE tissue sections in order to find out which cell types present in the tissue express particular proteins of interest. For this purpose I want to use CellProfiler to identify all the cells based on DAPI (primary object recognition) and define the cell boundaries/shape based on cell surface/morphology (e.g. membrane protein) staining (secondary object recognition). The latter should help me to define the “cell body” in order to actually determine which cells are “positive” for certain proteins (i.e. make CellProfiler search for positive staining within the secondary object masks).
Now I am facing the following challenges when looking for an appropriate way to stain the “cell bodies”:

  • CellMask stains (Thermo Fisher) nicely stain the cell surface, but they are lipid-binding reagents that do not work on FFPE.
  • Phalloidin-based staining (F-actin) is an option for FFPE, but I am worried that the cell borders are not defined with good enough contrast (especially in regions with high cell densities), which will make it very hard to define the cell borders (secondary masks) in CellProfiler (I might be wrong though, so please let me know if you had good experience with it…)
  • A good option would be to stain cell membrane proteins, but since I do not want to stain separate membrane proteins for each possible cell type present in the tissue, I am looking for a very ubiquitous membrane protein, which is expressed on virtually every cell.

Now to the actual question: Is anyone aware of (or has anyone ever successfully used antibodies against) a very unspecific/ubiquitous membrane protein that could serve as a cell type-independent cell surface marker to determine cell shape? Or are there other alternatives to the above-mentioned that are compatible with FFPE?

Otherwise I would try phalloidin or, in the worst case, use CellProfiler’s “MeasureObjectNeighbors” module to define a specific radius around the nucleus as the “cell body”, which of course would be way more “dirty”.

I am looking forward to a good discussion with plenty of suggestions from more experienced people!

2 Likes

I share your concern for phalloidin. In my experience it is not a great way to see the entire cell.

I am using wheat germ agglutinin for this kind of thing (but in cell culture, not tissue). Similar to phalloidin, it can be bought conjugated to many different fluorophores. It binds to glycoproteins covering most (all?) mammalian cells.

I often use it on live cells, just a brief 10 min staining.

1 Like

There are plasma membrane specific dyes that bind certain lipids (not all) that can be used. There are also cholesterol probes…I’ve seen others use a caderin protein, their expression can differ across cell lines depending on tissue of origin, differentiation etc. The cadherins also have some cytoplasmic staining but it can typically be thresholded out so that it doesn’t impact the image processing. TAhe junctional protein ZO-1 is quite clean. Some people use mitochondria stains, which tend to fill the cytoplasm and extend will out to the periphery…hmm…or any cytoplasmic probe/stain which fills the cell… Any cell line where the cells partially overlap will present a problem…which, given you are in tissue, is tough…if your cells don’t touch, or only a small fraction of them touch (<2%), phase-contrast images can also sometimes be used, historically it is more challenging to find shapes/ROIs (cells) without fluorescence, but I’ve heard the tools for this have improved a lot. Good luck!

1 Like

Thanks for your suggestion! I was googling around for WGA in FFPE and it might be an option, although I am not sure how well-defined the cell borders will be in tissue. But lectins in general may be interesting: Concavalin-A or Sambucus Nigra Lectin (Vectorlabs) seem to be quite a nice alternative to WGA. The question how abundant the corresponding carbohydrate-targets are across cell types… Maybe someone else in the forum has some experience with lectins on FFPE. One could perhaps also combine different lectins and hope to end up with a somewhat complementary staining.

1 Like

Hi James, thanks for your input. The biggest problem is that membrane dyes do not work on FFPE since lipids are washed off as part of the tissue preparation. I think given the vast diversity of cadherins and thight junction proteins it is going to be very difficult to find a “ubiquitous” one that serves the purpose. Perhaps some cytoplasmic may work, provided that they create some contrast at the edge of the cells and at cell-cell contacts.

Not a wetlab person, apply grains of salt liberally. I ran into the same issue with FFPE and membrane staining a while back. They attempted to use phalloidin once I pointed out that the membrane stains wouldn’t work. It didn’t work well enough to be useful.

Could you use a variety of cell surface antibodies ($$$) from the same animal so that the same secondary antibody would stain a variety of cell surfaces? Is this even reasonable?

I tried what I think was wheat germ agglutinin (verifying now) for segmentation in Imaris, but the staining was too spotty to use as a surface in order to separate cells. It ended up looking more like a spiderweb, though this is likely cell type dependent. Note that this was an attempt at 3D segmentation with a 63x, so at lower resolutions or in your FFPE tissue sections it might work better!

1 Like

Thanks for sharing your experience with WGA. I would rather not like to use various antibodies to stain many different surface proteins, since I would prefer to use my fluorescence channels to study my actual proteins of interest ^^. As you might know, the only thing that is more limited than money are the fluorescence filter settings of a microscope…

Sure, that is why I recommended using the same secondary (same primary animal), you would only be using one color.

1 Like

Aaah, I see. Sorry, I did not immediately get it… This might actually be an elegant “work-around”.

2 Likes

Along with cadherins the catenins are another possibility. Beta-catenin can stain cell borders nicely depending on tissue type. Laminin is another possibility. I doubt there is one marker that will work for all cell types. What tissue are you targeting?

Hi johnmc,

Thanks for your suggestions and sorry for the late reply (I was on vacation…). I ordered labeled lectin (ConA), which binds alpha-linked mannose residues on glycosylated proteins and should therefore, at least in theory, stain more or less all cell types. If this does not work, I will think about other options (maybe a combination of cell surface markers or receptors as you suggested). I am targeting brain tumor tissue and therefore it is really difficult to anticipate all the different cell types (e.g. immune cells, healthy cells from surrounding brain tissue, …) that might be present in a given tissue. In addition, there is also heterogeneity in the expression of surface proteins within a given tumor cell population and between tumor tissues. This is the reason why I want to be as unspecific as possible.

Just saw this on twitter, as an example of WGA working. Maximum intensity projection of a Z-stack, I believe, though very clean cells and nice separation. In tissue, might be another story altogether. YMMV.

Hi All,

we tested Cellmask, ConA, WGA, and DiI on FFPE tissues (liver, small intestine, kidney). The most satisfactory staining pattern was in the liver. The gut and kidney were generally a lot messier, so your mileage will vary with the tissue of interest. Labeling was done for 20 min at either RT or 37C on 3 um sections. Good luck and let me know if you want more details or some data to play!

2 Likes

Thanks for your input!

In which buffer did you prepare your ConA staining solution? I just read that ConA is Ca2+ and Mn2+ dependent and therefore, buffers that bind double-positively charged ions (e.g. PBS) should be avoided. In addition: Did you perform a blocking step prior to ConA staining?
Thanks for your help!

Hmmm, I think we used PBS without calcium and magnesium and there was no blocking step that I can see in the notes. I will double check and let you know if I’m mistaken. Slides were deparaffinized in Propar, rehydrated in graded isopropyl alcohols down to 70%, stained with Hoechst and ConA (or other dyes) in PBS, washed with PBS and mounted.

I see that reference in the Molecular Probes manual. Vector labs recommend this buffer for labeling reaction: 10 mM HEPES, 0.15 M NaCl, pH 7.5, 0.01 mM Mn++, 0.1 mM Ca++, 0.08% sodium azide.
It will be interesting to see what kind of a difference it makes!