[JVGen]

Just found this post while searching google about Glycine quenching. It's quite annoying to see prior posts have been edited to remove information in anticipation of publication.

Have your results been published since? This thread would be useful, but all the good bits and pieces have been removed.

[kylini]

In bacteria I don't think you can get away with those short times. Takes time to pass through all the barriers.

What was your final quench concentration for Tris?


Quote from a this paper----Brent Sutherland, et al., "Utility of formaldehyde cross-linking and mass spectrometry in the study of protein–protein interactions" 2008 Journal of Mass Spectrometry, 43, 699

"The addition of the buffer, Tris, appears much more efficient in stopping formaldehyde cross-linking reactivity. Even at its lowest concentration (1.2- fold), the degree of modification is significantly lower than that obtained for glycine at the same concentration"


Glycine does work to a degree and maybe quite well in certain conditions but I don't think people check as to just HOW effective it is in quenching. Someone used it and then it became a protocol and then a standard in biology and when something doesn't work for slightly different condition your PI may say "He's not as good as this other kid". Anyway, Tris does seem to better quencher than glycine.

I can even kind of see how it became a trend... "Well formaldehyde cross links proteins and proteins are made of amino acids and we have a lot of glycine in our lab which is an amino acid and there's this chem paper that says this reaction works so let's go with that" -- Conditions, they do matter.

Regarding concentration: 1% formaldehyde is really 333 mM. I calculate however many mols is present in my given final volume, multiply it by 2.5, and then add an appropriate amount of 3 M Tris in the neutral-basic range. 750 mM final concentration happens to be a decent ballpark for that number. That said, it's overkill and it does make spinning cells down require a ton of 50 ml conicals (unless you're adherent, then SCORE). Feel free to titrate down as desired.

Regarding that paper: yuuuuuup! Most of my literature searching involved crystallographic sources and, strangely, some industry whitepapers on embalming. That paper had the clearest "cite me" figure.

Regarding tradition: I do love how 2/3 of protocols are Young Lab clones and the remainder follow ENCODE. I'm not saying they're bad, but I do really wonder what those washes truly do.

Regarding blind assumptions and crosslinking length: sometimes, tradition isn't a bad thing. I assumed that lowering crosslinking times would "magically" improve signal:noise by eliminating non-specific binding. I was right! *NOTHING* but Pol II bound to chromatin after 30 sec of CH2O and ChIP-Seq. It was the cleanest Pol II dataset I've ever seen, but my other 20 multiplexed samples were very very devoid of useful information (thank goodness I only paid for 1 lane, hoping to add more for depth after checking the result). Increasing crosslinking time has since re-enabled capture of those "accessory" factors. The terrible quenching of glycine probably played a role in ChIP-Seq working for everyone else!

[TheSeqGeek]

I do suppose a good experimental explanation is in order!

For the limited number of lanes shown, I added a 3-fold molar excess of glycine or Tris directly to formaldehyde and mixed (a so-called "prequench" since if it works, formaldehyde should be totally dead). I then introduced that mixture to HeLa cells to crosslink (or not ideally), spun down, washed, and lysed the cells, and then sonicated. The resulting sheared chromatin samples were either heated to 65 overnight with RNase and Proteinase K to reverse crosslinks, or instead kept at 4 degrees (RNase-treated only) to keep crosslinks intact. The resulting samples were phenol:chloroform:IAA extracted and ethanol precipitated (meaning that the crosslink-reverse-minus lanes represent UNCROSSLINKED DNA ONLY while crosslink-reverse-plus lanes represent TOTAL DNA).

It's obvious the two Tris-prequenched lanes are identical. This is because crosslinking never occurred and, thus, sonication was very efficient (naked chromatin). Recovery was high in both lanes despite skipping crosslink reversal (implying no loss to phenol due to protein-DNA linking). Tris quenches formaldehyde.

On the other hand, the two glycine-prequenched lanes are drastically different. Sonication was very inefficient (crosslinked chromatin) in the de-CH2O-plus lane and only uncrosslinked DNA (de-CH2O-minus) had good efficiency. Further, a huge amount of sample disappeared from the crosslink-reversal-minus lane compared to the total DNA lane. This is because crosslinking occurred despite prequenching and the resulting DNA-protein complexes were trapped in the phenol interphase. Glycine, thus fails to quench formaldehyde.

Take home message: if you want to control crosslinking time, use Tris. This will allow you to go *very short* with your times if desired (30 seconds, 1 min, etc.) and see differences in sonication and ChIP signal. However, I do not recommend such short times, since the reason ChIP works is longer crosslinking times! If you go that short, you'll only get Pol II and histone sequencing to work! (Yeah... oops)

In bacteria I don't think you can get away with those short times. Takes time to pass through all the barriers.

What was your final quench concentration for Tris?


Quote from a this paper----Brent Sutherland, et al., "Utility of formaldehyde cross-linking and mass spectrometry in the study of protein–protein interactions" 2008 Journal of Mass Spectrometry, 43, 699

"The addition of the buffer, Tris, appears much more efficient in stopping formaldehyde cross-linking reactivity. Even at its lowest concentration (1.2- fold), the degree of modification is significantly lower than that obtained for glycine at the same concentration"


Glycine does work to a degree and maybe quite well in certain conditions but I don't think people check as to just HOW effective it is in quenching. Someone used it and then it became a protocol and then a standard in biology and when something doesn't work for slightly different condition your PI may say "He's not as good as this other kid". Anyway, Tris does seem to better quencher than glycine.

I can even kind of see how it became a trend... "Well formaldehyde cross links proteins and proteins are made of amino acids and we have a lot of glycine in our lab which is an amino acid and there's this chem paper that says this reaction works so let's go with that" -- Conditions, they do matter.

[kylini]

I'm a bit late to the party but what are those gels showing exactly... The Tris looks the same in both conditions...

I am doing ChiP-seq now myself

I do suppose a good experimental explanation is in order!

For the limited number of lanes shown, I added a 3-fold molar excess of glycine or Tris directly to formaldehyde and mixed (a so-called "prequench" since if it works, formaldehyde should be totally dead). I then introduced that mixture to HeLa cells to crosslink (or not ideally), spun down, washed, and lysed the cells, and then sonicated. The resulting sheared chromatin samples were either heated to 65 overnight with RNase and Proteinase K to reverse crosslinks, or instead kept at 4 degrees (RNase-treated only) to keep crosslinks intact. The resulting samples were phenol:chloroform:IAA extracted and ethanol precipitated (meaning that the crosslink-reverse-minus lanes represent UNCROSSLINKED DNA ONLY while crosslink-reverse-plus lanes represent TOTAL DNA).

It's obvious the two Tris-prequenched lanes are identical. This is because crosslinking never occurred and, thus, sonication was very efficient (naked chromatin). Recovery was high in both lanes despite skipping crosslink reversal (implying no loss to phenol due to protein-DNA linking). Tris quenches formaldehyde.

On the other hand, the two glycine-prequenched lanes are drastically different. Sonication was very inefficient (crosslinked chromatin) in the de-CH2O-plus lane and only uncrosslinked DNA (de-CH2O-minus) had good efficiency. Further, a huge amount of sample disappeared from the crosslink-reversal-minus lane compared to the total DNA lane. This is because crosslinking occurred despite prequenching and the resulting DNA-protein complexes were trapped in the phenol interphase. Glycine, thus fails to quench formaldehyde.

Take home message: if you want to control crosslinking time, use Tris. This will allow you to go *very short* with your times if desired (30 seconds, 1 min, etc.) and see differences in sonication and ChIP signal. However, I do not recommend such short times, since the reason ChIP works is longer crosslinking times! If you go that short, you'll only get Pol II and histone sequencing to work! (Yeah... oops)

[TheSeqGeek]

Fun update!

Here's the edited experimental setup:

17.6 million HeLa cells were treated for 30 seconds with 1% methanol-free formaldehyde premixed for 3 minutes with a 3-fold molar excess of either glycine or Tris.

Here's the result:



In conclusion, glycine does not quench formaldehyde but Tris does.

[omitted]



I think this settles my problem. If you have quenching issues, give Tris a try! It's certainly been used in the literature; this just confirms that it works.

I'm a bit late to the party but what are those gels showing exactly... The Tris looks the same in both conditions...

I am doing ChiP-seq now myself

[kylini]

Did you check the pH of your glycine stock versus Tris?

Both were titrated to pH 7. Glycine was almost there to begin with.

[omitted]

[austinso]

Did you check the pH of your glycine stock versus Tris?

[kylini]

Isn't there serum in the cell culture media that you're adding formaldehyde to?

Yup. 10% I believe. I don't see the media "unquenching" the formaldehyde that's been pre-mixed with glycine, however. If anything, I would expect the media to help quench, but it doesn't do enough.

I can't stress this enough. I still have sonication issues. However, this settles my quenching issue.

[Joann]

Isn't there serum in the cell culture media that you're adding formaldehyde to?

[kylini]

Fun update!

Here's the edited experimental setup:

17.6 million HeLa cells were treated for 30 seconds with 1% methanol-free formaldehyde premixed for 3 minutes with a 3-fold molar excess of either glycine or Tris.

Here's the result:



In conclusion, glycine does not quench formaldehyde but Tris does.

[omitted]

I think this settles my problem. If you have quenching issues, give Tris a try! It's certainly been used in the literature; this just confirms that it works.

[kylini]

Well, I don't know what career-stage you are at, but spending two years figuring out sonication and ChIP-Seq is far, far, far too long in my opinion.

Don't I know it! I've definitely been pursuing other experiments on the side and this is definitely not the only thing my lab does. It'd just be nice to actually do these experiments in house and not rely on collaborators. Notably, I already have samples sequencing despite the sonication (horray for size selection after library prep) so this isn't holding me up. Other fun deep-sequencing projects are in the works too!

It's just that this glycine first experimental result goes completely against established protocols. I'm sure the number of molecules of formaldehyde vs. the number of cells (ignoring volume) is part of the problem. I was curious if anyone else knew what could be going on.

[theduke]

I can't drop the line of thinking that glycine isn't quenching formaldehyde in these conditions because I have reproducible, empirical evidence showing it doesn't and it explains my sonication failures despite two years of optimization using protocols directly from the labs of Richard Young and the ENCODE Consortium. That's why I think that there are other factors at play, such as the number of molecules (not the concentration) per cell.

Well, I don't know what career-stage you are at, but spending two years figuring out sonication and ChIP-Seq is far, far, far too long in my opinion. At the very least, I would have moved on to MNase a long time ago in your situation. Actually a compelling argument can be made for the preferential use of MNase over sonication these days, but that is a different discussion.

I would start thinking about optimization experiments in small dishes and then work up from there. Have you done this? I don't like the large number of cells and excessive volumes you are using. I typically do my x-linking in 10 cm dishes with 5 ml media at RT. I usually do multiple dishes at the same time but do not combine them into one massive container.

ENCODE ChIP-Seq data is far from perfect!!

[kylini]

That is correct. [omitted]

The reason I'm so concerned about the quenching is that 1% = 10 g per liter. Since CH2O has a molecular weight of about 30, it follows that there are 30 g per mol. 10g/L times mol/30g = 333 mM. How can something that reacts 1:1 to stop formaldehyde from crosslinking further possibly work at a 333 mM to 125 mM ratio?

Again, I added 125 mM glycine FIRST before formaldehyde and the cells went right on to over-crosslink (all of the material was crosslinked). I then added 333 mM glycine FIRST before formaldehyde and the cells STILL went right on to over-crosslink (again, all material was crosslinked). If I'm crosslinking in media, there should be a bit of extra quenching power from its components. Despite this, the amount of glycine I'm adding doesn't actually quench.

If quenching doesn't actually happen, that means my cells are crosslinking while I aliquot them into 50 ml tubes (another 5-10 minutes) and while they pellet (another 5 min). Instead of crosslinking for 30 seconds, I'm really crosslinking for 30 minutes, hence the reason my sonication sucks.

I can't drop the line of thinking that glycine isn't quenching formaldehyde in these conditions because I have reproducible, empirical evidence showing it doesn't and it explains my sonication failures despite two years of optimization using protocols directly from the labs of Richard Young and the ENCODE Consortium. That's why I think that there are other factors at play, such as the number of molecules (not the concentration) per cell.

[theduke]

I don't think there is much doubt that glycine quenches formaldehyde. I would drop that line of thinking quickly if I was you. There is too much literature and too much evidence against you and it can't all be wrong. There is also no doubt that fixing time per se influences shearing efficiency. Something else is going on with your experiments. Either your glycine is off/concentration is wrong or something is up with your formaldehyde or your technique. It looks to me that you are massively over-crosslinking.

Generally, I have 10^8 cells per 200 ml, up to 6*10^8 cells per 1.2 L. I grow HeLa cells at a concentration of 5*10^5 cells/ml until I'm ready and then dump them into a vessel already containing the requisite 16% formaldehyde (final conc. 1%).

This is where I think your problems are. Something doesn't smell right. That is a ****-ton of cells and a lot of volume. Are you doing the quenching etc in suspension?

[kylini]

This is not the first time I have seen shearing like this using the Covaris. I think the glycine thing is a "red herring" here and I would suspect your shearing methodology is the problem. Have you tried any other sonicators? I would hope so after two years.

I use a Fisher Sonic Dismembrator 550 and, recently, have confirmed that I am able to use the Covaris. Both sonicators result in the same profile and both were used for quenching experiments.

My sonication certainly isn't yet optimal, but it still doesn't change the result that adding 125 mM or 333 mM glycine BEFORE adding formaldehyde yields the SAME crosslinking efficiency as CH2O exposure without it. I have no reason to trust that glycine is actually quenching formaldehyde and this is why I need input.