Small Adapter Dimer Can Cause Big Troubles!

[pmiguel]

Over time we apparently became too complacent about adapter dimers. We just did a library check Miseq run on a bunch of TruSeq DNA libraries that looked like this (10x dilution) :



Okay, the "128" bp adapter dimer peak looked a little more prominent than we would like, but we did not really give the trace a second look.

I should have looked at the molarity! Actually, the smaller adapter dimer amplicons must have clustered preferentially, because this sample clocked in at 50% adapter dimer! Needless to say that pool will be getting re-size selected (ampure).

Here is the 25-6-25 cycle miseq %Base profile:



Yes, you can read the proximal adapter sequences right off the %base profile..

--
Phillip

[GeGnome]

Yes, we've had some like this too! And you can read the sequence off like a Sanger run. You're right...you have to look at the molar amounts of your adapter dimer to your library. We've allowed a small peak in, but only if our library peak was much larger.

I guess this is how we live and learn. I think we just over clustered half of our flow cell and under clustered (400K) the other half. Part of the learning curve...

Thanks for sharing...I find these posts helpful...

[lorendarith]

So what do you do then? Do you take the molarity of the dimer into account and overload the lane?

[GeGnome]

What we do is look at the relative molarity of the adapter dimer in relation to the library and just make a decision from there to redo the library, to get rid of the dimer, or to just proceed with sequencing. If the adapter peak is small, thin, and just a blip compared to the library peak, then we've done ok just going forward.

However, in many cases our next step is exome enrichment which makes the adapter dimers less of an issue since they should be sort of filtered out. IMO, if you have a dimer, the best approach is to do a quick low-end size selection and just get rid of the peak. If the adapter dimer is huge, then there is probably a problem with the library itself and it may make the most sense to redo the prep.

[lorendarith]

We've been starting to see them since we've been using the Ampure XP beads for the size selection instead of the gel. They are always present when we use the beads and we can't get rid of them no matter how many times we size select. The beads were properly calibrated and the upper cut-off works perfectly.

As you said, in some cases we can ignore them as the molarity is much smaller compared to the library, but when we do long insert size genomic libraries when the end product is present in small amounts the peak is often bigger and we aren't sure whether to take the molarity of the dimer into account when loading it on the machine. The price and amount of DNA for long insert size libraries is quite hefty so we are happy when there is even a small amount of end product. Starting again is not quite the option.

Also, we have a colleague who quite often has these high primer peaks even though she uses the same protocol. We've been joking it is becoming kind of her personal signature. Here's a pic. Maybe there is something wrong with the beads or not and they just preferentially bind the smaller fragments... I have no idea, but currently the only way to properly get rid of them is size selecting on a gel :/

[TonyBrooks]

The way to reduce them is to put more DNA in at the beginning. We typically see this when people make libraries following mis-quantification of input (Nanodrop just isn't good enough). This means the 20:1 ratio of adatper to DNA is out promoting dimerisation of adapters.
Also, we found reducing PCR cycles can help a bit. Adapter dimer will be preferentially amplified during enrichment PCR and on the flowcell surface during cluster gen. Due to reaction kinetics, smaller PCR pretty much always work better than smaller ones.

[eab]

In my experience, a 1:1 ratio of SPRI beads to sample removes them almost entirely while retaining >90% of 250-bp fragments
Pipetting must be done very carefully, of course. A few extra mcl of buffer/beads in the form of a droplet on the outside of the pipet tip can make a big difference.
Worth considering whether any given batch of beads needs to be titrated using a MW ladder, in case batches differ.

[pmiguel]

The way to reduce them is to put more DNA in at the beginning. We typically see this when people make libraries following mis-quantification of input (Nanodrop just isn't good enough). This means the 20:1 ratio of adatper to DNA is out promoting dimerisation of adapters.
Also, we found reducing PCR cycles can help a bit. Adapter dimer will be preferentially amplified during enrichment PCR and on the flowcell surface during cluster gen. Due to reaction kinetics, smaller PCR pretty much always work better than smaller ones.

(I presume you meant " smaller PCR pretty much always work better than larger ones.")

I completely agree with your take on this. I would just add, that getting rid of the adapter dimer prior to denaturing the sample (during, for example, enrichment PCR) helps. I would speculate that after the first denaturation, you would expect some of the adapter dimer to anneal to full length library molecules, where it could hide from most purification methods.

A little probably exchanges off and reforms double-stranded adapter dimer peak each cycle of ampure. Just enough to taunt you.

Doing a couple of cycles of ampure will deplete it to an extent.

--
Phillip

[kmcarr]

Have you guys seen this posted just recently on the Illumina web site (need iCom account):

Improved TruSeq RNA v2 sample preparation

Sep 25 2012

You may have noticed that we recently modified the TruSeq RNA v2 protocol to include a heat inactivation procedure immediately following the Adenylate 3’ Ends step. This modification involves a short incubation performed at 70°C for 5 minutes followed by a 4°C hold. We have found that this simple workflow change can reduce the presence of adapter dimers and concatemers, improving the quality of libraries generated from the kit.

You are welcome to immediately add this simple workflow change to the kits you currently have in stock.

For more information on the additional heat inactivation step please see the “Adenylate 3’ Ends” section of the updated TruSeq RNA v2 user guide, available for download here: https://icom.illumina.com/download/s...Y0u7rCyqv5EX-A

We have a vexatious relationship with adapter dimers; low input RNA seems to be a major contributing factor.

[pmiguel]

Yes, we changed the appropriate thermal cycler profile to add the heat inactivation. Not sure what this is in aid of. Kill the polymerase, less chance of carry-over into the ligation step where it might A-tail a blunt adapter?

--
Phillip

[captainentropy]

Yes, we changed the appropriate thermal cycler profile to add the heat inactivation. Not sure what this is in aid of. Kill the polymerase, less chance of carry-over into the ligation step where it might A-tail a blunt adapter?

--
Phillip

I don't use the RNA-seq kits but are the adapters blunt on one end? It would have to have the A-overhang on one end for the ligation to work.

Nevertheless, I'm glad I found this post. We'll try it with our ChIP-seq libraries. Sometimes some people in the lab get adapters dimers that just won't go away.

I agree with TonyBrooks about the DNA amount issue (more DNA = fewer unligated, and mischievous, adapters left behind) but I also agree with lorendarith, some people's technique is the issue. We had a tech in the lab that had this issue a lot. In an attempt to resolve this issue multiple times we did libraries side-by-side and using the same DNA, at the same time, on the same cycler, etc., I didn't get the dimers while the tech did. I couldn't identify what was being done differently. Something subtle. Weird.

Just to add to what pmiguel showed, in this attachment you can see that small, ostensibly insignificant (relative to the rest of the trace), peak at ~128bp (adapter dimer) constituted disproportionally significant amounts to the sequencing. Like TonyBrooks also said, smaller fragments amplify better.

But, we are transitioning as much as we can to protocols using hairpin adapters. No more adapter problems with them.

[pmiguel]

I don't use the RNA-seq kits but are the adapters blunt on one end? It would have to have the A-overhang on one end for the ligation to work.

The adapters have a "T" overhang on the ligatable side.

Nevertheless, I'm glad I found this post. We'll try it with our ChIP-seq libraries. Sometimes some people in the lab get adapters dimers that just won't go away.

I agree with TonyBrooks about the DNA amount issue (more DNA = fewer unligated, and mischievous, adapters left behind) but I also agree with lorendarith, some people's technique is the issue. We had a tech in the lab that had this issue a lot. In an attempt to resolve this issue multiple times we did libraries side-by-side and using the same DNA, at the same time, on the same cycler, etc., I didn't get the dimers while the tech did. I couldn't identify what was being done differently. Something subtle. Weird.

Just to add to what pmiguel showed, in this attachment you can see that small, ostensibly insignificant (relative to the rest of the trace), peak at ~128bp (adapter dimer) constituted disproportionally significant amounts to the sequencing. Like TonyBrooks also said, smaller fragments amplify better.

Actually your numbers roughly fit expectation. It's number of molecules that determine cluster density, not mass.

But, we are transitioning as much as we can to protocols using hairpin adapters. No more adapter problems with them.

Hairpin? You mean single long oligos that form a stem loop? Seems like they could still seed primer dimers...

--
Phillip

[captainentropy]

The adapters have a "T" overhang on the ligatable side.

Exactly. So we're back to your original question of what the point of the 70 degree step is. But hey, if it works, then that's a problem solved. But I want to know why. hahaha.

Hairpin? You mean single long oligos that form a stem loop? Seems like they could still seed primer dimers...

Yes, aka stem loop. Reducing/eliminating adapter dimers is actually a feature of these kinds of adapters. So far we haven't seen any when using these kits.

[pmiguel]

Exactly. So we're back to your original question of what the point of the 70 degree step is. But hey, if it works, then that's a problem solved. But I want to know why. hahaha.

Well, if your sample has any really short duplex DNA fragments, those might be ligated between two adapters. But a 70 degree step would melt these apart, removing them as possible reactants.

70 degrees also probably denatures the polymerase that is used to add the 3' A overhangs. That might help.

Yes, aka stem loop. Reducing/eliminating adapter dimers is actually a feature of these kinds of adapters. So far we haven't seen any when using these kits.

Okay, but why? Illumina adapters are "Y adapters" -- only the 10 bases adjacent to the 3' T overhang are duplex. The rest is single stranded.

--
Phillip

[captainentropy]

I like the idea that small fragments would be denatured, but wouldn't they reanneal? Plus they'd have to probably be at least ~40 bp in length since small fragments should have been removed in the first purification step (which is essentially a PCR cleanup). But then they might show up on the bioanlayzer trace as peaks at 140 bp, 150 bp, 160 bp, etc. But I never see those peaks...

I can't speak to the Klenow fragment used in the Illumina kit but I use the one from NEB and for heat inactivation their instructions say:

Heat Inactivation
75°C for 20 min

But, maybe 70 C for 5 min is enough...

As for the stem-loop adapters in the kits available now (at least for the ThruPlex kit from Rubicon that we use) they are a DNA-RNA hybrid. After ligation to the DNA fragments they are inactivated and thus can't generate amplifiable dimers.

Here's the patent for the kit that explains why the adapters don't form http://www.google.com/patents/US7803550