Mitochondria contamination

[yimmieg]

Hi,

We are trying out the new ATAC-seq protocol:

http://www.nature.com/nmeth/journal/vaop/ncurrent/full/nmeth.2688.html

and we are seeing a lot of mitochondria contamination. Does anyone have an ideas to how to get rid of this? It's > 60% of all reads. My thoughts are some old school biology (making biotin probes for the roughly 16k bases)?

~Jimmie

[doc2r]

Hi Jimmie,

I am also seeing the same issue with contamination of chrM reads. My correspondence with the authors tells me they see the same thing in their data and that they are trying to improve the methods. At the moment we are trying to wash the "nuclei" more than once.

What i've also noticed is that there the PCR dups rate is pretty high. Do you see this?

[Jubs]

Hi Jimmie and doc2r,

I'm about to try this method also and I'm glad I found this thread in time. Did you have any success in cleaning up for mtDNA? Or did the authors gave any suggestions?

For now I'm considering a double-wash on the nuclei and qPCR for mtDNA sequences to see how enriched they are in relation to the rest. That should give an idea on how many washes are needed to get rid of most mtDNA and, most importantly, how much of the good stuff I lose when washing more than once.

When analyzing their original data I also found the 60% of the reads mapping to chrM, which reduces a lot the amount of reads that are actually usable. Funny they don't mention this in the paper...

Our sequencing facility experts said that use of biotin probes for getting rid of ribosomal dna would not be so straightforward because smaller genDNA fragments would also get pulled down... but I never tried this before so I wouldn't know how they would actually behave.

Thanks!

[doc2r]

Likewise we have not had much success removing Mitochondrial DNA.
PCR dups are still a big issue. We are trying different tagmentation times to see if we are over tagmenting. This could be leading to small fragments which get preferentially amplified... hence higher dup rates..

[doc2r]

Jubs,

When you processed the authors data, did you check to see if there were any PCR duplicates?

[wjyzidane]

Jubs,

When you processed the authors data, did you check to see if there were any PCR duplicates?

Hi doc2r，

the duplicates rate for authors data is about 40%~60% in our analysis.
Actually I am also trying to ATAC-seq but the chromatin seems undertagmented so I am wondering whether you are still use 37 for 30min and how many cells you use for the tagmentation.

[wjyzidane]

Hi Jimmie and doc2r,

I'm about to try this method also and I'm glad I found this thread in time. Did you have any success in cleaning up for mtDNA? Or did the authors gave any suggestions?

For now I'm considering a double-wash on the nuclei and qPCR for mtDNA sequences to see how enriched they are in relation to the rest. That should give an idea on how many washes are needed to get rid of most mtDNA and, most importantly, how much of the good stuff I lose when washing more than once.

When analyzing their original data I also found the 60% of the reads mapping to chrM, which reduces a lot the amount of reads that are actually usable. Funny they don't mention this in the paper...

Our sequencing facility experts said that use of biotin probes for getting rid of ribosomal dna would not be so straightforward because smaller genDNA fragments would also get pulled down... but I never tried this before so I wouldn't know how they would actually behave.

Thanks!

Hi Jubs,

I am wondering how many cells you are using for atac-seq, because if the cell number is limited, double wash would lead to loss of cell easily.

Jingyi

[Sir Ranzelot]

Hi everyone,
I have the same problem with up to 60% of the reads mapping to mitochondria. In my case, almost all of the PCR duplicates come from this source- reads mapping to the "normal" chromosomes do not have many duplicates, meaning when I could remove mitoDNA I would also get rid of PCR duplicates.
In general ATAC-seq worked straightforward with PBMCs. There, the libraries did show kind of a nucleosome "ladder" from the tagmentation. With different cell lines I do not see such a nice "ladder" as in the libraries prepared from PBMC ATAC-seq. I tried different cell numbers and lysis buffers, but results do not improve much. Was anyone more successful in removing mitochondria reads and making nice libraries?
Thanks a lot

[bysanimadhu]

hi,
Can anyone have updated protocol for ATAC-seq. It seems all of you have the problem with mitochondrial DNA contamination? Did you able to get rid of this in the later experiments. May I know which cells you are using?

[wen yuan]

Hi bysanimadhu,

Do you have more detailed protocol for ATAC-seq? I can not fully understand their protocol in their nature methods.

hi,
Can anyone have updated protocol for ATAC-seq. It seems all of you have the problem with mitochondrial DNA contamination? Did you able to get rid of this in the later experiments. May I know which cells you are using?

[bysanimadhu]

Hi, please join in the ATAC-Seq forum. Its maintaining by the Greenleaf lab. You can find many things including the detailed protocol. Here is the link for the forum. https://sites.google.com/site/atacseqpublic/.

Hi bysanimadhu,

Do you have more detailed protocol for ATAC-seq? I can not fully understand their protocol in their nature methods.

[wen yuan]

Hi bysanimadhu, Thanks a lot. I have applied their forum yesterday. But until now no response yet. Anyway thank you very much.

Hi, please join in the ATAC-Seq forum. Its maintaining by the Greenleaf lab. You can find many things including the detailed protocol. Here is the link for the forum. https://sites.google.com/site/atacseqpublic/.

[Jubs]

Hi everyone,
Well, I have finally managed to get my 1st ATAC sequenced and analysed. I'm not working with cell lines, so I have yet another black box here which is dealing with frozen tissue - cell/nuclei integrity, quality of the tissue, time which it had to wait until it was frozen, defreezing, etc.

Anyway, after playing with different buffers/spinning/washing/incubation time/enzyme concentration, I gave up and basically followed the published protocol.
Actually, I found that 0,1% detergent in the lysis buffer was too much, so I went down to 0,01% (the background was too high otherwise, meaning there was an uniform smear in the gel and I wasn't getting any qPCR enrichment for my controls).

After analysing the seqs, I saw that mtDNA contamination was really low (1%), but the noise was also high, i.e., some of the "peaks" were faded out in the background - or at least not so sharp as DNAse peaks, which I was using for comparison.
(Also, PCR duplicates were really low, so that wasn't a problem for me.)

I have the feeling that the less mtDNA in the sample, the higher the noise. Maybe if one is too harsh when isolating nuclei, that ends up killing all mitochondria, the nuclei also get disrupted somehow and then more chromatin is accessible for the enzymes... don't know, just an idea.

For the next round I think I'll just repeat the procedure and, if I get nice qPCR enrichments, I just sequence it deeper to try to improve my signal-noise ratio.

That's my experience up to now. If anyone has also tried it out with tissue rather than cell lines I would like to hear

Best,
Ju

[yimmieg]

hi ju,

cool to see that you have reduced mtDNA with changing lysis conditions. did you ever do a technical replicate with .01% and .1%? it would be nice to quantitate the background noise at different lysis conditions. i don't think mtDNA contamination per se would change your signal to noise ratio but the changing lysis conditions might.

as for primary cells, we have had some experience working with blood cell types and our results are generally similar to what the original ATAC-seq paper reported in T cells. so as far as we know, the protocol works just fine in primary cells.

~jimmie

[Jubs]

Hi Jimmie,

No, I have only sequenced the library that gave best results. I agree it would be nice to have technical replicates with the different detergent concentrations, but, you know, money... :P

I measured mtDNA by qPCR, and the only 'useful' information I have is the obvious: more detergent, less mtDNA (plus, with no detergent at all, mtDNA goes up to the roof).

And in agreement with what I said in my previous reply, the more mtDNA, the higher the enrichments of regions of interest. So, in samples with high mtDNA, the fold enrichments of regions of interest are higher than in samples with less mtDNA.

Cheers!