help with "colocalization thresholds"

[Itsaso Garcia]

Hey everyone!
i´m pretty new in using ImageJ, and sometimes i have doubts than don´t
meet an answer so easily...
i´m trying to quantify the colocalization of two proteins in cultured
cells. I mark each protein with a fluorofore (red and green) and i
take confocal micrographies. Since i´m going to repeat the experiment
several times (obviously), i´m very careful in keeping always the same
adquisition parameters in the microscope (voltage of PMT). Afterwards,
i correct the background with "background substraction", the same
rolling ball in the cases. That gave reasonable results in some
experiments, but in the last one i got a very dim intensity in the
channel for red colour. It is so that, when quantifying (i run the
plugin "colocalization thresholds" for that), the threshold set is
below zero, or above the maximum intensity value in any pixel of the
image!!
Is it neccesary to keep constant those parameters inter-
experimentally? Or, can i define them for each experiment?

I don´t know if i explained my situation properly so one can
understand it. Sorry, i´m not familiar with these words and concepts.
Thanks! i hope someone can help me! i don´t want to waste time doing
non-correct analysis!!
Best regards,
Itsaso.
 

********************************
Itsaso Garcia-Arcos
Department of Physiology,
University of the Basque Country Medical School
UPV/EHU, Spain.
********************************

[Juan Carlos Marvizon]

I'm glad that your brought this up. I am also new to ImageJ, and I'm trying
to use it to measure colocalization, with not much luck.

I work with spinal cord sections double- and triple-labeled with several
antibodies. I take confocal images with a Leica microscope. This is what I
got so far...

- I can't get ImageJ (v1.37f, with the WCIF load of plugins) to open my
Leica stacks. It does open the individual TIFF files.
- I have used the Intensity Correlation Analysis (ICA) and Colocalization
Threshold (CT) plugins on matched pairs of optical sections (individual TIFF
fiels).
- I ran several controls. Positive controls are two antibodies (different
species) labeling the same epitope (neurokinin 1 receptors or the peptide
CGRP). This should give maximal colocalization. Negative controls are two
antibodies that are not expected or observed to colocalize.
- Of all the parameters produced by the plugins, only the Pearson's
Correlation Coefficient (Rr) seems realistic. Mander's Overlap Coefficient
(R) and Mander's Correlation Coefficients (M1 and M2) are always high, for
both positive and negative controls.
- The histograms produced by the plugins correspond well with the expected
levels of colocalization, but they seem to have little relationship with R,
M1 and M2.
- I am also using another program, Imaris Colocalization, to analyze the
same data. I got a free trial license until the end of May. Imaris loaded
the whole Leica stacks without problem, and analized them together. It
produced much more realistic parameter values than ImageJ.
- Imaris also has an automatic thresholding feature for colocalization.
However, this feature failed for some data stacks. For the other data, it
tend to give colocalization vaues that are too high (even for negative
controls). I decided not to use it, and set the threshold by eye.

I don't mean to advertise for Imaris. I much rather not spend the money and
use ImageJ. Imaris is quite expensive! But this is where I am right now...

To adress your question, I think that background subtraction is redundant
with thereshold. Why eliminate some dark pixels if you are going to exclude
them from the analysis later on, by setting a threshold? Also, the important
Rr parameter is independent of thereshold, and my instinct is to obtain it
for the raw date. By doing background subtraction you are biasing the data
already. If you are doing things right on the confocal, your background
should be very low, anyway.

According with my experience with Imaris, threshold should be set "by eye",
and a bit high. I use a pixel value of 60-80. Automatic threshold can be
problematic. If you use it, you should run some controls, both positive and
negative.

I hope this helps you... If somebody reads this and can help with my
problems with ImageJ, I'll be grateful.

Buena suerte!

Juan Carlos Marvizon