a way to quantify protein localization based on its fluorescence intensity distribution

Authors: Liang Cai
Published: Creative Commons Attribution-Noncommercial 3.0 License
Version: 9
Last edited: 05 Oct 2011
Exported: 23 Nov 2011
Original URL: http://knol.google.com/k/-/-/3ez005w9gduud/4

Prepare files for the analysis

The file EdgeRatio.txt is the ImageJ macro written for this specific purpose, which is attached at the end of this page. This macro can also be download from my homepage - code section.

Before using the macro, one value in EdgeRatio.txt need to be adjusted. Looking into the file, and change the 787 to an appropriate value. With the demo-RGB-file.tif capture condition, 787 pixel equals to 50 micron.
// 787 pix == 50 um
xValues = ((i - realpair) * steppix) * 50 / 787;

Here is an example image demo-RGB-file.tif. (It was in the tif format. Due to the size limit, I uploaded it as png format. Please convert it back to RGB tif file in ImageJ.)

Edge detection process

Open any file in RGB format in ImageJ, and run the macro EdgeRatio.txt by Plugins->Macros->Run...

As shown above, a window pops up. Step pixel is the step the mask dilating or shrinking in each measurement. Total pixel distance is how far inside the cell will be measured. With the demo-RGB-file.tif, input 36 pixel will measure about 2.29 micron inside the cell.

OK to proceed.

As shown above, ImageJ will open many windows automatically. So, this macro cannot be run in batch mode. The polygon selections tool is enabled. Outline the area of interested. Make sure enough space is left inside and outside of the cell. The yellow line show the ROI for lamellipodial measure, which covers the major protrusion on the right and avoid the ruffle in the middle. Convert the selection to mask by Edit->Selection->Create Mask.

The wand tool is selected now. Using Image->Adjust->Threshold to set the image as a binary image (usually, single press "Auto" will do the trick; no need to "Apply" or "Set"; occasionally, drag the bar to get it precise). Use the wand tool to select the outline of the cell, as shown below.

To archive a good outline for the cell edge, which is the critical step in the whole process, adjusting the bars in the Threshold windows helps. Only the region within the ROI needs to be precise. It is OK to use the Paintbrush tool to fill some gaps or smooth the regions outside the ROI to help the selection. Using the wand tool NOT the freehand tool to generate the outline, which minimizes the bias.

As shown above, from this point, the macro starts running by itself, and many windows are open and close. DO NOT INTERRUPT! NO MOUSE MOVING OR KEYBOARD CLICK. The progression bar in ImageJ shows the status of each dilating/shrinking operation, which takes most of the calculation time.

After a while, all the images are close and only the Results window left in ImageJ. This result table is stored as CVS format with a name end with .xls, which means it can be opened with Excel etc. No more operation is need in ImageJ. The content of Results window will be flushed when the macro starts running.

Check the result file

In the xls file, RGB/rgb means the red, green and blue. 0/1/2 = r/g/b. Distance is the real distance in micron, with "-" means inside the cell. Intensity is the raw value of the channel, which is averaged over the edge within ROI. SemOfInt is the standard error of the mean of the raw value, which is calculated from the thousands point of the edge. PercInt is the percentage value of the intensity, with the maximal raw value as 1 and the minimal raw value as 0. The PercInt is the most consistent value between different cells. Group different cells and a new SEM can be calculated to show the variance between cells. B2g-Ratio is (raw value of blue channel) / (raw value of red channel), etc. Exact the same microscopy setting is required for the ratio and raw value meaningful between different cells.

ImageJ macro for the analysis

The macro is now in github EdgeRatio.ijm.

It fix a NaN bug. The script from my homepage doesn't have this fix.

© 2012 Endogenous - CaiLog. blogged by cail.cn - 35cd076bdd09fcbdfe7ca24eb8f93900

For fluorescence imaging, the halogen lamp was used with two filter cubes, one with excitation (BP520-540HQ) and emission (BP555-600HQ) filters for observing mKO2 fluorescence, and the other with excitation (470DF35) and emission (510WB40) filters for observing mAG fluorescence.

mKO is from http://www.biochemj.org/bj/381/0307/3810307.pdf

mAG is from http://www.jbc.org/content/278/36/34167.full.pdf

It is very safe to treat mAG as EGFP. But there is a problem with mKO. It is orange but not red. It can be excited a little by the 488 laser, which is used to excite FITC/EGFP! I don't think it is possible to overcome the problem, especially when the red is strong and the green is weak. Do you?

© 2012 Endogenous - CaiLog. blogged by cail.cn - 35cd076bdd09fcbdfe7ca24eb8f93900

There is a workshop going on right now. I sit through half of it and realized that all I need to know is what functions she used (Macros.zip). Because, I use ImageJ macro a lot (one example).

Here are my notes from her code.

• ImageJ 2 is coming, combined effect of BioFormat, Fiji, ImageJ
• in macro window, highlight the rows then command+r / ctrl+r will only execute part of the code, perfect for debugging
• start from recording your own macro (menu: macro - record), and understand its code
• nSlices for the slice number of an image stack (z stack or time lapse)

• exist1 = File.isDirectory(pathToSaveTo + "max");
  if (exist1 != 1) {
  	File.makeDirectory(pathToSaveTo + "max");
  }

• for (i = 1; i <= n; i ++) { // looping code}

• for (n = 1; n <= nSlices; n ++){
  	setSlice(n); // go to slice n
  	run("Measure");
  }

• for (n = 1; n <= nSlices; n ++) {
  	setSlice(n);
  	getStatistics(area, mean, min, max, stdev);
  	if (mean > maxMean) {
  		maxMean = mean;	//if it is larger than maxMean, then mark this slice as the new maxMean
  		frameToDup = n;
  	}
  }

• orig = getImageID;
  
  newImage("cropped", "16-bit Black", 300, 300, nSlices);
  final = getImageID;
  
  setBatchMode(true);
  for (n = 1; n <= nSlices; n ++) {
  	selectImage(orig);
  	setSlice(n);
  	run("Copy");
  	selectImage(final);
  	setSlice(n);
  	run("Paste");
  }
  setBatchMode(false);
  
  //find the frame with the maximum intensity, go to that frame and reset brightness/contrast
  maxmax =0;
  setBatchMode(true);
  for (n = 1; n <= nSlices; n ++) {
  	setSlice(n);
  	getStatistics(area, mean, min, max);
  	if (max > maxmax) {
  		maxmax = max;
  		nofmax = n;
  	}
  }
  setBatchMode(false);
  
  // now go to slice with the maxmax and use that to reset intensities in the stack
  setSlice(nofmax);
  run("Brightness/Contrast...");
  resetMinAndMax();
  
  //convert the new stack into 8-bit, which is required for 3D volume projection
  selectImage(final);
  run("8-bit");
  //run the 3D projection viewer. Do a "record macro" to get your settings right. Make sure to enter the appropriate xy to z pixel ratio
  //for example, 200nm steps with 61nm/pixel in x,y is a ratio of 3.28
  run("3D Project...", "projection=[Brightest Point] axis=Y-Axis slice=3.28 initial=0 total=360 rotation=9 lower=1 upper=255 opacity=0 surface=100 interior=50 interpolate");

• resetMinAndMax(), With 16-bit and 32-bit images, resets the minimum and maximum displayed pixel values (display range) to be the same as the current image's minimum and maximum pixel values. With 8-bit images, sets the display range to 0-255. With RGB images, does nothing. I usually do "Enhance Contrast", which actually will have over-expose problem, not ideal. I need to check for a 0-1000 range 16-bit image, what's the final 8-bit value will be using resetMinAndMax() - 8bit, or using "Enhance Contrast" - 8 bit

Now, you are ready to dig deeper - familiar with all built-in macro functions and use them!

© 2012 Endogenous - CaiLog. blogged by cail.cn - 35cd076bdd09fcbdfe7ca24eb8f93900

I came across a short report. In its supplemental figures, I found the following image. It very nicely summarizes what I should do to detect spots from noise images. From step a to h, less than 20 lines matlab script will do all for you! Wow~~

DOI: 10.1038/nmeth1024

© 2012 Endogenous - CaiLog. blogged by cail.cn - 35cd076bdd09fcbdfe7ca24eb8f93900

I developed it because:

1. I don't like to manually input reference; I want the program do it for me - automatically record all the paper I read or scanned. Using the browser plugin, I easily achieved it.
2. I want to access it whenever/wherever I need. Store in the cloud and use a browser to access it is the best way! It is a pain to keep an updated database cross multiple computers. Might be easier if using Dropbox, but still, have to install some proprietary software and dropbox... Everything gets much easier when everything is in the CLOUD.

So, www.pubmeder.com born and starts serving me since last October. I have over 3000 entries in the cloud.

What if you already have a big collection of reference on your computer? How easy is it to transfer them to the cloud?

Super EASY!

If you alreay have a collection of PubMed ID (PMID for short), visit http://www.pubmeder.com/enter using your login. You will find "Batch Enter by PMID". Copy, paste and submit - done!

But, what if you only have the collection in some software, but not the PMID list? Here are some scripts you can use. First, export all your reference as text file, using BibTex or RIS format. In the case of EndNote, you want to use "BibTeX Export" or "RefMan (RIS) Export". Second, examine the exported file to check how PMID is wrapped in the text. In the case of "RefMan (RIS) Export" from EndNote, it is wrapped as "AN - 123456789", because in EndNote, PMID is saved as the "Accession Number" (short name "AN"). Third, after installing Perl, save the relative script as run.pl in the same folder where exported reference file locates. Execute it, input the file name, and all the PMID will be extracted and saved to result.txt. Enjoy!

For PMID wrapped as "AN - 123456789"

#!/usr/bin/perl
print "Please Enter the File Name for Extract : ";
my $fname = <>;
chop($fname);
open(FH, $fname);
open(RESULT,">result.txt");
my <a href="http://twitter.com/data" target="_blank" class="litwitter">@data</a> = <FH>;
foreach $line (@data) {
  $line =~ s/^\s*an\s*-\s*{(\d+)}.+/$1,/i;
  if ($line =~ m/^\d+,$/) {
    print RESULT $line;
  }
}
close RESULT;
close FH;

For PMID wrapped as "PMID = 123456789"

#!/usr/bin/perl
print "Please Enter the File Name for Extract : ";
my $fname = <>;
chop($fname);
open(FH, $fname);
open(RESULT,">result.txt");
my <a href="http://twitter.com/data" target="_blank" class="litwitter">@data</a> = <FH>;
foreach $line (@data) {
  $line =~ s/^\s*pmid\s=\s{(\d+)}.+/$1,/i;
  if ($line =~ m/^\d+,$/) {
    print RESULT $line;
  }
}
close RESULT;
close FH;

Oh well. If you still feel it is a lot of work, do comment below. I could easily make a webpage to do all the extraction for you, when I have a little free time~~

© 2012 Endogenous - CaiLog. blogged by cail.cn - 35cd076bdd09fcbdfe7ca24eb8f93900

This study shows that the transcription factor Glis1, which is highly enriched in unfertilized eggs and one-cell-stage embryos, promotes iPSC generation effectively and specifically by activating multiple pro-reprogramming pathways. Glis1 might thus be a link between reprogramming during iPSC generation and reprogramming after nuclear transfer. Furthermore, iPSCs generated by OSK and Glis1 did not cause a marked increase in mortality of chimaeric mice, although this did occur with iPSCs generated by Oct3/4, Sox2, Glis1 and Myc (Supplementary Fig. 10) and with iPSCs generated by OSK and Myc, as reported previously 17. The identification of Glis1 might therefore be beneficial for future applications of iPSC technology.

Very cool! Got one from 1400 ORF!

And, this blog post from Knoepfler Lab is very helpful in understanding the significance of this paper.

Are Glis1 iPS cells safer? While Glis1 did not apparently increase mortality in mice derived from Glis1-produced iPS cells, it remains unclear if Glis1 might have an oncogenic function. In fact it would be surprising if it didn’t given that not just Myc, but essentially all pluripotency factors are oncogenic or highly expressed in malignant tumors.

If Glis1 turns up Myc expression, could that make cells behave in a more tumorigenic manner? It’s unclear, but it is a concern. Even transiently high levels of Myc can lead to genomic instability and oncogenic transformation.

Is Glis1 expressed at high levels or amplified in human cancers? The paper does not comment on this and a brief database search did not find any connection, but there is very little if any data on this.

What is the normal function of Glis1 in early development? Who knows, but could be to keep N-Myc (and possibly L-Myc) levels high in the very early embryo.

Glis1

© 2012 Endogenous - CaiLog. blogged by cail.cn - 35cd076bdd09fcbdfe7ca24eb8f93900

Huang W., Ghisletti S., Saijo K., Gandhi M., Aouadi M., Tesz G.J., Zhang D.X., Yao J., Czech M.P., Goode B.L., Rosenfeld M.G., Glass C.K.

Coronin 2A mediates actin-dependent de-repression of inflammatory response genes.

Nature, 2011 Feb 17;470(7334):414-8

Toll-like receptors (TLRs) function as initiators of inflammation through their ability to sense pathogen-associated molecular patterns and products of tissue damage. Transcriptional activation of many TLR-responsive genes requires an initial de-repression step in which nuclear receptor co-repressor (NCoR) complexes are actively removed from the promoters of target genes to relieve basal repression. Ligand-dependent SUMOylation of liver X receptors (LXRs) has been found to suppress TLR4-induced transcription potently by preventing the NCoR clearance step, but the underlying mechanisms remain enigmatic. Here we provide evidence that coronin 2A (CORO2A), a component of the NCoR complex of previously unknown function, mediates TLR-induced NCoR turnover by a mechanism involving interaction with oligomeric nuclear actin. SUMOylated LXRs block NCoR turnover by binding to a conserved SUMO2/SUMO3-interaction motif in CORO2A and preventing actin recruitment. Intriguingly, the LXR transrepression pathway can itself be inactivated by inflammatory signals that induce calcium/calmodulin-dependent protein kinase IIγ (CaMKIIγ)-dependent phosphorylation of LXRs, leading to their deSUMOylation by the SUMO protease SENP3 and release from CORO2A. These findings uncover a CORO2A-actin-dependent mechanism for the de-repression of inflammatory response genes that can be differentially regulated by phosphorylation and by nuclear receptor signalling pathways that control immunity and homeostasis. [PMID: 21331046]

Below are my comments to some of their data:

• Fig.1b: no control for cytoplasmic staining - I could get similar image for any protein by just moving up the focus plane
• No mutant protein stability estimate - I bet that their delta N mutant is not stable. Not sure about SIMmut, K11A/R13A; which might be fine.
• Fig.5b: use IP Actin to show the actin binding activity - Never seen it in the literature! In the buffer with EGTA, all actin filaments are disassembled. If it is such a robust G-actin binding protein, why is it not in cytoplasmic?
• Why actin is recruited to the promoters? Again, based on their experimental details, it is G-actin binds to the promoters. But they claim a mechanism involving interaction with oligomeric nuclear actin.
• Why using 10 uM NEM in their buffer for co-IP and ChIP? NEM, a blocker of vesicular transport, reacts with cysteine containing proteins? Since all vesicles were broken in the buffer with NP-40, NEM will reacting with all Cysteine containing proteins in the lysate? Is the result reliable?
• No comments on NCoR clearance, SUMO, LXR, CaMKII, etc. I am not an expert of those experiments.

Because of the limitation of 140 characters in a tweet, the actually posted comments are:

PMID: 21331046 @NatureNews: No comments on NCoR clearance, SUMO, LXR, CaMKII, etc. I am not an expert of those experiments.

PMID: 21331046 @NatureNews: ... not block vesicular transport. NEM will react with all Cysteines on proteins in lysate! Are results reliable?

PMID: 21331046 @NatureNews: Why using 10 uM NEM in the buffer for co-IP and ChIP? Since all vesicles were broken in the buffer with NP-40...

PMID: 21331046 @NatureNews: Based on exp details, it is G-actin binds to promoters. But the claim is a mechanism involving oligomeric actin

PMID: 21331046 @NatureNews: In buffer with EGTA, all actin filaments are disassembled. Such robust G-actin binding, why not in cytoplasmic?

PMID: 21331046 @NatureNews: Fig.5b - use IP Actin to show actin binding activity - Never seen in literatures! With EGTA, filaments all gone.

PMID: 21331046 @NatureNews: No protein stability estimate - I bet their delta N mutant is not stable. Not sure about SIMmut, K11A/R13A

PMID: 21331046 @NatureNews: Fig.1b no control for cytoplasmic staining - I could get the image for any protein by moving up the focus plane

Let's see~~

© 2012 Endogenous - CaiLog. blogged by cail.cn - 35cd076bdd09fcbdfe7ca24eb8f93900