Help identifying membrane from imperfectly labeled image


#1

Hello,

I’m a very new user of CellProfiler and I’m having trouble figuring out a strategy for segmenting cell membranes labeled with a genetically encoded (i.e. not externally stained) fluorophore. I have high-mag images which contain a few cells labeled with:

BFP-H2B (very tightly confined to nucleus)
RFP-FarnesylationTag (shows significant enrichment on cell membrane, but there is weak cytosolic fluorescence and seems to clump in the ER as well)
GFP-ProteinX (protein of interest tagged with GFP).

I am interested in tracking how well ProteinX localizes to the membrane, which I think is conditional on a certain stimulus. I can identify cells using IdentifyPrimary on the nuclei, and I was then planning to use the RFP image as a guide for co-localization to see where the membrane is, and then see what fraction of the GFP image is within that mask. However, the localization of RFP to the membrane isn’t perfect, and I don’t want the cytoplasm and ER to be inlcuded in the mask. I can get rid of the cytoplasm by thresholding. Is there some sort of topology-based filter that I could use for eliminating the internal cluster at the ER?

I’ve uploaded the image files:

Outputs.zip (928.8 KB)

Thanks for any help!


#2

If you just want to see how much is in the membrane, I’d do it like this

-IDPrimary for the nuclei as you have been doing
-IDSecondary using your nuclei and the RFP channel to find the whole cell
-ExpandOrShrinkObjects to shrink your Cell objects just a little bit (start by trying 2-3 pixels)
-IDTertiary to subtract your shrunken Cell objects from the original full-sized Cells- this’ll give you an object that’s just a 2-3 pixel border around the edge of the cell that you can call Membrane, and you can measure your amount of GFP in this Membrane object.

Does that make sense? Give it a try and see how it goes; if not well, feel free to upload your original pipeline so that we can try to adjust or tinker with it.