When is it recommended to normalize protein fluorescence to nuclear fluorescence?

Hi,
I am staining for proteins in the cell nucleus and would like some help with specific questions and the intuition behind when it is appropriate to normalize the single cell protein intensities to the single cell nuclear (DAPI/Hoescht) fluorescence. Either if someone can directly address the questions below or redirect me to some good reading material, that would be much appreciated.

When measuring the mean intensity per pixel in an object (cell), I understand that in addition to differences in the amounts of the molecule of interest, there are many other factors that can influence the intensity, including

  • Changes in nuclear volume (e.g. from cell compaction).
  • Cells at slightly different focal z-planes.
  • Cells growing on top of each other so that their nuclei overlap and seem brighter than they are.
  • Cells at different stages of cell cycle (at least this affects total DAPI)

Correctly accounting for these confounding factors would yield more accurate image quantification, and I have a few specific questions regarding how to go about this:

  1. Are these the most common confounding factors that are intended to be accounted for when normalizing protein to nuclear intensity?
  2. I have seen papers where they simply divide the intensity for the protein of interest with the nuclear intensity. Is there a good reference that shows that the nuclear and protein intensities co-varies linearly when the potentially confounding factors change?
  3. Are there instances when it is recommended to not normalize to nucelar content?
  4. To what degree does confocal slices get rid of the need for normalizing nuclei at different Z-position?
  5. Is the mean DNA intensity expected to be the same between all cells and not vary as the total DNA intensity does when cells progress through the cell cycle?
  6. If the system under study allows for it, would the best approach simply be to ensure that all cells are evenly spread out as a near perfect monolayer and avoid the need for any nuclear normalization?

As an example or the intuition, I include a couple of images.

The left shows nuclear intensity and the right shows a protein localized to the nucleus. The red ring indicates an area that is higher in nuclear intensity than the surrounding cells (not including the small, very bright dividing cells). The same area does not show any tendency of also having higher protein expression than the surrounding cells. I understand that the cells with the highest nuclear intensity will not necessarily be the cells with the highest protein intensity (then we wouldn’t need to measure both in the first place). However, in the dimmest cells expressing little or no protein, I would expect to see a weak variation in their intensity similar to the variation in nuclear intensity, which would suggested that the nuclear intensity is a good indicator of which cells are affected by confounding factors in the protein stain.

Thanks for reading my somewhat lengthy post.

Bump, does anyone have any pointers to where I can find more information for this?

Hello,

I can give you some general advice.

Mean vs sum intensities

  • the amount of protein per cell is measured by integrated (sum=total) intensity
  • the average local concentration of protein is measured by mean intensity (total/volume)

It often is a huge biological difference whether you measure the mean or the sum intensity!!!
You have to think about it carefully; i’ve seen many projects go wrong here.

Example: H2B-mCherry in HeLa cells during cell cycle

I found that during interphase mean intensity of H2B-mCherry stays mostly constant, indicating that the nucleus grows in size as much as there is new H2B added and thus the concentration stays more or less constant; but H2B-mCherry sum intensity increases because the amount increases. During mitosis it is the opposite: mean intensity increases because the DNA gets concentrated in the meta-phase plate; but sum intensity stays more or less constant because no new H2B-mCherry is produced (in fact that’s not really the case in the actual measurement but I rather think it is an optical artifact…).

Anyway, the point is that mean and sum really can behave very differently.

Confocal vs wide-field

  • mean intensities are typically only meaningful in confocal images, because the imaged volume per pixel is defined by the spatially limited confocal PSF. In wide-field microscopy the imaged volume per pixel is infinite and thus not well defined.

  • mean intensities are typically only meaningful in confocal images, because the imaged volume per pixel is defined by the PSF. In wide-field microscopy the imaged volume per pixel is infinite and thus not well defined.

  • total intensities per cell for (thin) tissue culture samples are typically better measured with wide-field microscopes. Because of the unlimited PSF you get the whole fluorescence of the cell in one shot, i.e., you don’t need z-stacks. For thick samples, where unwanted signal can be below or above your signal of interest you however need a confocal microscope.

Normalization

This is another critical choice that depends very much on the specific biological questions and has to be carefully chosen to match the biological question. Different normalizations report different biological measurements. So there really is no general advice, but you have to think for every project very carefully how you normalize.

Literature

I hope this helps somewhat. And please let me know if you find a publication about these very important questions, because, in fact, I was considering writing one if there is none.

2 Likes

Thank you for your response Christian! Sorry for not getting back to you earlier, it seems like I am not getting notifications on this topic for some reason…

I am not sure that I understand your point about why mean intensities would only be useful in confocal images. Wouldn’t the PSF smear out an object to nearby pixels in two dimensions as well?

I tend to use the mean/median intensity, mostly because that is what I have been recommended from senior lab members to compensate for cells changing in the area of cells spreading out differently and then being captured in a 2d image. I think it would also be more robust to strong background, which might add to the integrated intensity. I would be interested in reading some good comparison on this if there is one available.

Please let me know if you find/write something regarding normalization of protein stains to nuclear stains. I understand that different normalizations can report different biological measurements, but I am still curious about which normaliations can be used to compensate for imaging artifacts from e.g. cells in different focal planes.

Cheers!

Hello,
The main thing you need to understand is that the biological meaning of mean vs. integrated intensities is different (see my example “H2B-mCherry in HeLa cells during cell cycle” above). In many cases it could really be about publishing a paper with a different title, using the one or the other. Using my example as a guideline, would you be able to say what you need to measure to answer your biological question? (The background affects mean and integrated in the same way.)

Yes, I agree that the biological meaning is different between the two, although I would not necessary say that I understand exactly which biological questions are answered more effectively by measuring the protein concentration (mean intensity) in the cell or the total abundance (integrated intensity).

Naturally, protein concentration is the most akin to qualitatively looking for bright fluorescence in an image. I would also think this is the most indicative of functional changes in the cell. Increased concentrations of proteins would lead to higher availability for participation in reactions locally within the cell, whereas a higher total amount of the protein might not be mean that protein is actually highly available for reactions within the cell, unless these cellular reactions would be able to actively recruit proteins from the entire cell volume, rather than relying on local concentrations and collisions of proteins.