SEQanswers

SEQanswers (http://seqanswers.com/forums/index.php)
-   Sample Prep / Library Generation (http://seqanswers.com/forums/forumdisplay.php?f=25)
-   -   genomic DNA contamination before RiboZero and ScriptSeq (http://seqanswers.com/forums/showthread.php?t=79553)

arctan 12-11-2017 09:20 AM

genomic DNA contamination before RiboZero and ScriptSeq
 
This will be my first time making RNA-seq libraries. I have the RiboZero (for bacterial RNA) and ScriptSeq library prep kit.

I prepared bacterial total RNA and did on-column DNase treatment. NanoDrop 260:280 ratios are 2.0-2.2. The RiboZero protocol recommends quantifying RNA concentrations using Qubit, so I used the RNA Broad Range assay for that. For kicks, I also tried the dsDNA High Sensitivity assay on the same samples.

I am getting about 6-8% genomic DNA contamination compared to the concentration of RNA, but naively, I was expecting much less.

Has anyone done similar DNA vs RNA Qubit measurements for their samples for library prep, and what ranges do you get?

Also, how bad is my 6-8% contamination for RiboZero and subsequent library prep, since the genomic DNA will still be present after rRNA depletion?

Since a RNA purification step is required after RiboZero, perhaps I can proceed with RiboZero first, then do an additional DNase treatment immediately after, and then proceed with the RNA purification step? Or should I not worry about the 6-8% of DNA?

Any recommendations or advice on how to proceed would be much appreciated! I am completely new to this...

pmiguel 12-12-2017 09:23 AM

Hi arctan,
You might want to take an aliquot of your sample, heat it to 70 oC for a few minutes to denature it's secondary structure, snap cool and then retest for the presence of genomic DNA.

Although the Qubit dsDNA High Sensitivity fluor detects primarily double stranded polynucleotides, I do not think it distinguishes between double-strand DNA and double-strand RNA. So any secondary structure in your RNA can read as DNA.

To the extent your RNA has long stretches of double stranded regions, the 70 oC treatment will not be sufficient to denature it.

Another alternative would be to take an aliquot of your RNA and treat it with a very effective RNAse, maybe if you can find a mixture that can degrade both ssRNA and dsRNA. Then check this aliquot for DNA.

--
Phillip

jteeee2 12-12-2017 10:55 AM

That material will be fine for a total RNA Seq approach. I wouldn't get concerned unless that value is above 10%. Also agree with pmiguel that the Qubit value can be misleading.

arctan 12-13-2017 10:58 AM

Thank you both for the helpful replies!

arctan 06-01-2018 08:25 AM

Just wanted to follow up on this... Based on my recent experience, I think if you are doing RiboZero depletion (as opposed to Poly A selection, not an option for microbes), the library prep may be much more sensitive to genomic DNA contamination, since contaminating gDNA will remain after rRNA depletion, but should be selected against during poly A selection.

I recently had Total RNA samples with good 280:260 and 260:230 NanoDrop ratios which were also quantified by Qubit RNA BR and by Bioanalyzer. I made a bunch of directional RNA-seq libraries after Ribo-Zero depletion. QC'ed the libraries by Tapestation and Qubit.

After aligning the reads to the reference genome, one of the libraries showed decent directionality, i.e. most of the reads are on the same strand as the CDS. But for the remainder of the libraries, they showed absolutely lousy directionality, i.e. 50% of the reads are on the same strand as the CDS, and 50% are on the opposite strand.

I am actually very surprised, because this happened even with some RNA samples with RIN values >9 and there were no obvious genomic DNA peaks on the Bioanalyzer. I can understand how a little gDNA contamination might lead to a higher background of reads (i.e. in intergenic regions), but in my case, it seemed like the invisible contaminating gDNA ended up dominating my sequencing run. Does this surprise anyone else (especially those with more experience with RiboZero and RNA-seq)?

I will now have to treat my future RNA samples with a lot more attention to gDNA contamination.

pmiguel 06-04-2018 08:58 AM

Since Ribo-Zero depleted RNA will include non-poly A+ messages, then it isn't clear to me that they would be expected to be on the same strand as the cDNA. Certainly it is possible that various sorts of non-polyadenylated messages would derived from introns.

--
Phillip

arctan 06-04-2018 11:42 AM

Thanks for your reply. Since I am working with bacterial RNA, splicing and introns aren't applicable. I agree with you though for eukaryotes, there can be many reads in introns in Ribo-Zero depleted RNA.

For example, https://www.nature.com/articles/nbt.2972/figures/5 is from a paper which did (among other things) a comparison between poly A-selected and rRNA-depleted libraries, and the main difference was the increased number of reads in introns.

However, since I am not working in a species with introns, I think it is more likely that contaminating genomic DNA is the problem. I am doing some tests with additional DNase treatments.

pmiguel 06-04-2018 12:07 PM

Okay, your minus strand reads may come from genomic DNA. But you have to treat DNA pretty roughly to fragment it down to a size (less than 500 bp) that could be amplified by PCR after ligation of adapters.

Stranded library protocols typically will include Actinomycin D to prevent and DNA-templated 1st strand synthesis.

What are the implications of this? Without Actinomycin D, your DNA could denature during the step that is intended to remove secondary structures from RNA and then primers could anneal that reverse transcriptase could extend. But with Actinomycin D, that should not happen, right? So you can't get much (or any) faux 1st strand synthesis off DNA, which means there won't be any faux 1st strand template to create faux 2nd strand.

So you really not only need to have DNA contaminating your RNA prep, but you need the DNA to be degraded to pretty small sizes for it to end up in your library.

Not saying it can't happen. Just that it doesn't seem likely to me.

--
Phillip

nucacidhunter 06-06-2018 02:09 AM

Relatively short DNA fragments can result from incomplete DNase activity or degradation during extraction and denatured to ssDNA at elevated temperature used for linearizing RNA. Two possible mechanisms can result in synthesis of antisense sequencing competent library fragments from ssDNA:

1- Reverse transcriptase used in reaction can also use ssDNA as substrate for synthesis of a complementary strand which will go through legitimate cDNA reactions

2- Degraded or incompletely synthesised “Terminal Tagging Oligos” with missing blocking agent at 3’ end will be extended on ssDNA template. If there is left over cDNA primer from previous step (cDNA synthesis) then it can be extended as well giving rise to fragments flanked by terminal priming tags.

arctan 06-07-2018 12:30 PM

Thank you both for the replies!

As an initial test, I acquired some Turbo DNase and treated an RIN >9 total RNA sample. I also had a positive control DNA-only sample to make sure the DNase is working. I then quantified by Qubit dsDNA HS and RNA HS.

Qubit dsDNA HS on the DNA-only control showed that DNase treatment worked, i.e. the Qubit signal dropped to background levels.

Qubit dsDNA HS on the total RNA showed a signficant DNA signal, but there was _no difference_ with or without DNase.

As suggested a while back by pmiguel, I further treated the total RNA with RNase A and T1, and then remeasured Qubit dsDNA HS. Now, the signal went way down. It went down by 17-fold in the no DNase sample, and to background levels in the DNase-treated one. In other words, the majority of the Qubit dsDNA HS signal in the total RNA was indeed due to non-specific binding of the fluor to the RNA.

The RNase digestion then eliminates this non-specific signal and now Qubit dsDNA HS now gives a better estimate of the concentration of contaminating gDNA in the total RNA prep.

It was about 1.5 to 2%. I think this is significant (at least for ScriptSeq), as described above by nucacidhunter, especially if the contaminating stuff is smaller, due to incomplete on-column DNase digestion in the original RNA prep.

In addition, during Ribo-Zero depletion (which is what I am using), the contaminating gDNA will survive and now it will be vastly more than 1-2% of the sample, especially if I have depleted 90-95% of the RNA.

I'll run a small pilot where I do an additional in-solution Turbo DNase treatment before a library prep and see if that solves the problem of non-directional reads.

pmiguel 06-08-2018 05:11 AM

Thanks for posting your results.
I suspect that most of your dsDNA flourimetric signal comes from double-stranded RNA--ie, RNA folded into secondary structures. My guess is that the fluor used does not distinguish between dsDNA, dsRNA and RNA/DNA heteroduplex.

--
Phillip


All times are GMT -8. The time now is 05:55 PM.

Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2018, vBulletin Solutions, Inc.