Specificty of Histone mod Abs and controls

[burza]

Hello,

I hope I'm not doubling posts, I couldn't really find any threads on my exact issue.

OK. Let's assume that I want to ChIP genome wide for H3K4me1. I was thinking of best way of checking the specificity of binding of these kind of Ab. I had one idea: to do another ChIP in parallel for unmethylated version of the same epitope. Then my questions are:

1. Are there Abs available for methylated and unmethylated histones based on the very same peptide?

2. If there are, can I assume that K4me1 occurs at both H3 in a particular nucleosome? In other words, if there is enrichment for K4me1, does it mean that I won't see a signal for an unmethylated form or it is very well possible that one H3 can be monomethylated for K4 and the other one not?

3. What are your suggestions for QC of histone mod Abs?


Thank you,
Greg

[mudshark]

Hi Greg,

1) nope. would not make sense anyway as there are many other modifications possible on the same peptide (K4me2 e.g.) that potentially prevents binding of the 'unmodified 'AB. so, the pattern is not at all expected to be the inverse.

2) nope, this you cannot assume at all. you can also not assume that all the nucleosomes at this position are methylated in all cells of your samples.

3) use at least two different antibodies from two different sources and check for the overlap of enrichment. don't get too much excited about the so-called ChIP-grade of ABs. pure marketing trick

[burza]

Thanks a lot for your answer!

Greg

[GKM]

Hello,

I hope I'm not doubling posts, I couldn't really find any threads on my exact issue.

OK. Let's assume that I want to ChIP genome wide for H3K4me1. I was thinking of best way of checking the specificity of binding of these kind of Ab. I had one idea: to do another ChIP in parallel for unmethylated version of the same epitope. Then my questions are:

1. Are there Abs available for methylated and unmethylated histones based on the very same peptide?

Most likely yes. But that's not the right experiment to do, as was posted above

2. If there are, can I assume that K4me1 occurs at both H3 in a particular nucleosome? In other words, if there is enrichment for K4me1, does it mean that I won't see a signal for an unmethylated form or it is very well possible that one H3 can be monomethylated for K4 and the other one not?

It was answered above too

3. What are your suggestions for QC of histone mod Abs?

The problem with histone modification antibodies has been that most of them are polyclonals (although I see supposedly good monoclonals slowly appearing on the market one by one), and as a result, there is a lot of variation from lot to lot. Especially true for the heterochromatin marks, as the sequence context is ARKS for both H3K9 and H3K27, as a result H3K9me3 antibodies are notorious as being cross-reactive with H3K27me3 (and in my experience, with H4K20me3 too). The various H3K4 methylation states tend to be more consistent, but one has to always check. The way it is typically done is with the peptides that were used to generate the antibodies (abcam conveniently sells them all), usually by dot blot, although any assay that can measure the relative binding of an antibody to the modified, unmodified, and suspected cross-reacting peptides can be used.

[burza]

Thanks for your input GKM.
In our lab we've been also wondering why almost all of histone mod Abs are polyclonal. Our conclusion mainly was that probably having a polyclonal Ab will enhance your signal as there are going to be many slightly different Abs that can access the antigen better (and can substitute for one another) depending on different chromatin conformation that can prohibit efficient binding (e.g. due to not optimal crosslinking) and in case of monoclonal enrichment might be lost completely. Of course down side of this is a possible unspecificity of polyclonals.

[GKM]

They are polyclonals because it's more difficult to make monoclonals, although once you have good monoclonals you're set forever. I am not an insider in the antibody-making business so I don't know why there aren't more monoclonals out there, but hopefully, they will eventually exist for everything. As of now, you have to check every lot, at the expense of a lot of time and effort.

Which most people don't do, as they aren't really aware of the issue, which means that probably there is quite a lot of data out there that isn't really what it's advertised to be.

[burza]

But it is hard to believe that Ab companies do not put more effort into making reliable 'ChIP grade' monoclonals. I would think that there are some serious problems with obtaining good histone mod monoclonals.

What is also strange that every now and then I see information on Abcam website saying that they're out of stock for a particular Ab because the can't obtain good enough quality. Does that mean that all the publication that have used this Ab are rubbish or maybe just that the new lot came out not the way the wanted...

I agree as to the data quality out there at the moment.

[GKM]

You are absolutely right that making monoclonals for these is hard.

For the polyclonals, Abcam are supposed to be doing internal checks of each lot and they probably do that. But they won't give you the data, so you can't trust them 100% about it. The ones you are talking about were probably so bad that they simply couldn't have them on the market.

[mudshark]

monoclonals usually have a much lower IP efficiency than polyclonals. furthermore you always have to remember that histone tails are heavily modified on many different residues which might generate occlusion effects on the primary epitope. that is a clear advantage of polyclonals which might contain a subspecies of ABs that is less affected. but polyclonals are subject to the same problem (e.g. S10ph close to K9 prevents most (all) of the polyclonal ABs to recognize methylated K9).

fundamental problem of the polyclonals is the limit of serum. immunization with exactly the same peptide lead to different sera. therefore differernt batches of polyclonal sera might have totally different specificities. and sometimes the companies are simply not able the re-generate a serum of equal quality. that is probably one reason for why ABs are taken off the catalogue .

it is indeed very difficult (i would say almost impossible) to do a good QC on ABs. Customers and companies however frequently estimate the quality based on the "I see what I expect to see" principle, just look at the abcam website. that's simply entering a vicious circle. peptide competitions are limited by the choice of competing peptides. Ideally one should check all modification states of all peptides for cross reactivity. one step in the right direction would be the test on peptide arrays: http://www.activemotif.com/catalog/667.html or http://wpl054.bio.med.uni-muenchen.de/HISMAD/

my personal experience: tri-methylated lysine antibodies are often highly cross reactive with other peptides harboring a Kme3. H3K9me3 - H3K27me3 is the historical example. but also H3K4me3 ABs often strongly cross react with K9me3 and K27me3 and K36me3.. and vice versa. One should never forget that companies want to make money selling antibodies and scientist want to publish. this situation is not the best starting point for raising good antibodies.

[captainentropy]

In answer to the general question of "why don't companies put more effort into making better antibodies", I'm gonna be a bit cynical and say - they are a company. Their bottom line IS the bottom line and the less they put into production/QC the more money they make. And they can do what they do, or don't do, because, well, what else are you going to do? Make a custom antibody for everything you want? No. Up to the extent that we can shop elsewhere is reasonable, we are at their mercy.

That said, I was just at a meeting where I learned of this effort http://epigenome.ucsd.edu/antibodies.html This group has validated the epitope specificity for a large number of antibodies. A paper related to this will appear soon in Nature Structural and Molecular Biology ("Assessment of histone-modification antibody quality.")

In relation to burza's last comment I'd say "maybe!" Take a look at this http://epigenome.ucsd.edu/pdf_files/...K9Ac-Abcam.pdf If you bought this antibody expecting to get ChIP or Chip-seq or ChIP-chip data for H3K9ac you'd have nothing but crap, or at least good data for H3K56ac hahaha.

I had a long talk with the lab that has done/is doing this QC and it was a thought to do these blots for a fee (like a core lab) to validate the specificity of (histone mod) antibodies for the research community but it really should fall on the company to make sure their antibody detects/IPs what they say it does. Personally, I've invalidated a few antibodies from AbCam that they had to pull from the market. Did that happen before others used it and made erroneous conclusions? I'm not sure, but for certain people wasted their research money and time.

As for monoclonals, the difficult part is it takes a long time (and lots of money) to test enough clones to be sure you have a strong and specific hybridoma. That said, once you have it, it should be good for a long time. Polyclonals, as many commented, are better because of the multiple epitope conformations that are detected.

[GKM]

In relation to burza's last comment I'd say "maybe!" Take a look at this http://epigenome.ucsd.edu/pdf_files/...K9Ac-Abcam.pdf If you bought this antibody expecting to get ChIP or Chip-seq or ChIP-chip data for H3K9ac you'd have nothing but crap, or at least good data for H3K56ac hahaha.

And the one I last bought was actually H3K27ac... (the K9/K27 cross-reactivity seems to not be restricted purely to trimethylation)

And I had to throw about half of the 20-is antibodies I bought at the same time for similar reasons

I had a long talk with the lab that has done/is doing this QC and it was a thought to do these blots for a fee (like a core lab) to validate the specificity of (histone mod) antibodies for the research community but it really should fall on the company to make sure their antibody detects/IPs what they say it does. Personally, I've invalidated a few antibodies from AbCam that they had to pull from the market. Did that happen before others used it and made erroneous conclusions? I'm not sure, but for certain people wasted their research money and time.

From what I've heard about the company, the scientists in it are pushing for more QC, but they have to fight over it with the business people there.

As for monoclonals, the difficult part is it takes a long time (and lots of money) to test enough clones to be sure you have a strong and specific hybridoma. That said, once you have it, it should be good for a long time. Polyclonals, as many commented, are better because of the multiple epitope conformations that are detected.

The majority of our most reliable and strongest TF antibodies are monoclonals. Although you are in general right, for ChIP, the benefits of monoclonals greatly outweigh the lower IP efficiency you would get (assuming your polyclonal was good). This is especially true for histones which are the easiest things to ChIP and where the additional specificity provided is of greatest importance