too much mtDNA

[millers]

I'm new to the NGS field and have just started to analyze my first set of data. I have a single lane of paired-end DNA sequence with 8 samples pooled into one lane. My libraries were generated with the Trueseq kit (as per Illumina's instructions) and run on a HiSeq2000.

I've looked at several chromosomes and the data looks fine. So far so good. However, when I started working with the mtDNA I noticed that for 2 of the samples I have WAY more reads. Instead of ~12 reads per base, I have ~100 across the entire mtDNA.

The most obvious explanation I can think of is PCR bias or some sort of contamination, but all samples were prepared at the same time so I am not sure why only 2/8 of the samples are affected. Also the rest of the genome seems to not have this bias for those 2 samples. Am I missing something? Anyone else have this issue?

[SNPsaurus]

millers, there are usually many hundreds of mitochondria in the cell, so the copy number of the genome is much higher than the nuclear genome, and so the read coverage will be higher as well.

The sample variation might come from how you extracted and purified the genomic DNA. Some kits can lose more mtDNA than others, and some differences in input concentration or buffer make-up could cause a swing (but I'm just guessing).

[Genohub]

mtDNA

How did you isolate the mtDNA? Any checks to make sure you are not looking at both genomic and mtDNA? I recommend using control genomic and mtDNA primers to verify.

- Genohub

[BhariD]

presence of mtDNA in RNA-seq data?

I have similar experience with my human RNA-seq data for 6 samples. I am also finding lot of mt sequence in the libraries (~1% to 5%).

I would like to take this opportunity to ask a question. What percentage of mtDNA/mtRNA is expected in human DNA-seq or RNA-seq data that makes sense biologically? At what percent is it alarming or indication of library contamination? is that 5% or 20% for a given RNA-seq data? if I am only interested in nuclear human RNA/DNA, what's the best way to prevent or minimize mitochondrial sequences of organelle origin? On the other hand, what about the integrated mtDNA fragments in human genome..does anyone know what percent of mtDNA in integrated?

I would greatly appreciate your response!

[Genohub]

mtDNA in Total or polyA purified RNA

In my experience with mammalian cells, I typically see ~6-10% of mt_genomic DNA, mt_rRNA and mt_tRNA reads in total RNA preps. After polyA purification, this can increase to as much as 8-30% of your reads.

You can try isolating away the mt DNA/RNA before you start your prep, or design probes to remove this material. Ribo-Zero claims to remove mt_rRNA.

- Genohub

[sfbiologist]

My lab is having a similar problem. We sequence yeast genomic DNA and in some runs, the mtDNA reaches almost 40% of the reads. More typically it's in the 5-15% range.

In the past we had less than 1% mtDNA; this was when we were using our own homebrewed method of library preparation with our own adapters with inline barcodes. Things changed when we switched to using NextFlex kits with Illumina-style adapters. We have used two different methods of initial DNA isolation (Qiagen genomic DNA tip-500 or bead-beating in phenol/chloroform) -- both gave NO mtDNA in our old preps and both give mtDNA in the NextFlex preps. So at least in our case, it looks like the problem is arising during the NextFlex prep, I suspect during PCR (see below). Some mtDNA is expected and tolerable, but 40% is way too much...it's seriously reducing the amount of useful data we get from each run.

Millers: like you, we're seeing considerable variability from sample to sample. We're still trying to narrow down the factors involved, but I think the samples with the LEAST mtDNA are also the ones that had the highest DNA concentration just before the PCR step. We do a gel extraction to size our library, and the samples that have the most intense bands on the gel seem to be the ones with the least mtDNA in the final data. I'm curious whether you have noticed any patterns like this.

Genohub: you mentioned removing the mtDNA/RNA. How would you do this for DNA? The only method I'm familiar with is CsCl centrifugation, which I'd prefer to avoid!

[Genohub]

I've found that the mtDNA contamination is not really library kit specific, it has to do with the sample type and cleanup method used. Variations in contamination have to do with user handling or the sample being used. There are some sample prep kits out there to remove mtRNA.

To remove mtDNA, you could design biotin probes and pull down material with strep mag beads. Haven't heard about any available commercial kits that can do that.

- Genohub

[sfbiologist]

What's really striking is that in our old method, there were essentially NO mtDNA reads regardless of the method of DNA isolation, researcher doing the prep, etc. It would be great to figure out why. One major difference between our old method and the NextFlex kit is that we previously Qiaquick columns after every step; with NextFlex we use Ampure XP beads. So is there any reason to think that either:

a) Qiaquick columns select against yeast mtDNA?
or
b) Ampure XP beads enrich the yeast mtDNA?

Yeast mtDNA is supposedly large (~80 kb). So I'm kinda confused about why either case would be true, especially once the DNA is already sonicated.

[scooter]

Epicentre makes the Ribo-Zero Gold kit which is designed to remove both nuclear and mitochondrial rRNA contamination. I think they have both a mammalian and yeast version of the 'Gold' kits. I have heard they work quite well.