[edawad]

I did several PCR reactions to test (had enough pre-pcr sample to spare). I tried 0.5, 1, 2, and 4 ul of my 10 nm working soln primer with both 15 and 18 cycles, and only using 4 ul got rid of the smears (which would be 8x more than I usually use for ChIP). Perhaps I had more sample than usual, or my primer soln was too dilute.

Can't answer your question about the ramp step but sounds like something worth trying.

[jlove]

edawad, I also had that experience (re-PCR 2 cycles got rid of my HMW peaks). Did you feel that the ultimate easy solution was just to use more primer with every PCR reaction? One time I tried doubling my primer and didn't get any different result, but double may just not be enough. I'm hoping that I can instead add some sort of ramp step or something to my prep (like Phillip suggested) just to get the average size correct moving forward.

[edawad]

Thought I'd throw in my experiences with HMW smears. I was preparing libraries using the NEBnext kit and custom barcode PE adapters. The libraries had a normal peak at 350, then a consistent smear going from 500-10kb on bioanalyzer. I usually use small amts of primer (equiv to 0.5 ul 10 nm per 50 ul reaction), so after reading this thread I re-PCR'd with 4 ul of 10 nm primer which completely got rid of the HMW smear. So, +1 for limited primer concentration.

[jlove]

Hi Phillip,
I'm curious to know if you tried this slow ramp step and if it worked.

-Jen

How about for the last cycle doing a very slow ramp from 98 oC to 60 oC? That way the full length products will have lots of time to find each other.
--
Phillip

[ETHANol]

SYBR green on the qPCR machine.

Which reminds me, Kapa makes formulation of their Library amplification polymerase with SYBR green in it so you can monitor library amplification in real time and stop before you over-amplify. I know, I sound like a Kapa sales rep but it solves all these over amplification problems.

[pmiguel]

[...]

One more thing, the melting temp of the slower then expected running DNA fragments is really low and forms a sharp peak which says that they are only being held together by adaptor interaction.

[...]

Just wondering: what is your assay for melting temp? (I remember the old hyperchromicity plots from long ago. But have not seen one done for decades...)

Phillip

[ETHANol]

When I say one more round of PCR what I mean is I take the Qiagen minElute/Ampure purified library and add all the reagent PCR (primers and 2X Mastermix) heat at 98˚ C for 1 min, 60˚ C for 30 sec and 72 ˚C for 1.5 min.

One more thing, the melting temp of the slower then expected running DNA fragments is really low and forms a sharp peak which says that they are only being held together by adaptor interaction.

Daisy chains, bubbles, etc. it's all a little more complicated then I originally thought. The important thing is that the individual DNA molecules are not actually big (at least in my case) and will not interfere with sequencing or quantification by qPCR.

[pmiguel]

Phillip, great post and nice explanations. It popped up while I was writing mine (I'm doing an experiment so it took a while).

Hilarious. Looks like we are writing posts at the same time and converging to many of the same conclusions.

As with everything in biology, you have to be correct in that it's all four at work depending on the situation.

I'm pretty sure that what I see are 'bubbles' because: 1) if I take an amplified library and denature it and let it reanneal the products are found at the same larger size,

Interesting! We did one where we were seeing multimers and a high molecular weight smear for a 454 amplicon library. Heated the samples to denaturation temp, then turned off the thermal cycler. They slowly cooled to ambient. The multimers disappeared.

2) if I submit the library to one more cycle of PCR with fresh primers the fragments return to the correct size, and

Just fresh primers, or a whole new set of PCR reactants? Thing is, this doesn't really distinguish among the hypotheses. Daisy chaining probably becomes favorable under the same conditions that Bubble Product formation would -- high concentration of products with respect to primer concentrations.

And, your fix, another cycle, would work for the PCR overamplification hypothesis as well. (PCR favors shorter templates.)

3) If I reduce the number of cycles I don't get the larger products. In my case it all adds up to being 'bubbles'

The strange thing that kind of threw me off at first is that the larger products are exactly twice the size of the 'correct' sized products.

Well, that seems like an argument for Daisy Chaining. I guess a Bubble product might run exactly 2x the size of a normal double stranded product, but I can't think of any reason it would.

I started getting the larger products when I switch to the Kappa polymerase. Turns out you have to to a lot less cycles then with Phusion polymerase.

We just use the TruSeq reagents. Not sure which polymerase it uses.

--
Phillip

[pmiguel]

Does not fit all the data I have seen. I have seen cases where higher order multimers are formed. That is, a 1-mer peak (expected length), 2-mer peak (a little less than double the expected length), 3-mer, etc.

This would fit with a daisy chain hypothesis but not with a bubble hypothesis.

--
Phillip

Actually, now that I think about it more, the Daisy Chain and Bubble hypotheses could both support multimers beyond dimers. Probably the best way to distinguish between them would be to take half a sample you know is showing the phenomenon and treat that half with S1 nuclease. The run the treated and untreated on the bioanalyzer.

The Daisy Chain hypothesis posits double-stranded amplicons linked end-to end by adapter-ends annealing. So either S1 would do nothing, or slice off the inter-amplicon annealed ends. Resulting in slightly shorter than 1-mer peak.

The Bubble hypothesis would leave much of the multimeric amplicons single-stranded and allow them to be nearly completely degraded by S1. So the 1-mer peak would be the same, while the multimer peaks would be destroyed.

The "PCR Overamplification" hypothesis would result in no change after S1 nuclease digestion because the multimers would be completely double stranded.

My guess is that a mixture of all possible multimer formation modes are happening.

--
Phillip

[ETHANol]

My guess is that all 4 of these are at work in various situations. The length of the TruSeq adapters is probably the source of these issues. What about mitigation?

Phillip, great post and nice explanations. It popped up while I was writing mine (I'm doing an experiment so it took a while).

As with everything in biology, you have to be correct in that it's all four at work depending on the situation.

I'm pretty sure that what I see are 'bubbles' because: 1) if I take an amplified library and denature it and let it reanneal the products are found at the same larger size, 2) if I submit the library to one more cycle of PCR with fresh primers the fragments return to the correct size, and 3) If I reduce the number of cycles I don't get the larger products. In my case it all adds up to being 'bubbles'

The strange thing that kind of threw me off at first is that the larger products are exactly twice the size of the 'correct' sized products.

I started getting the larger products when I switch to the Kappa polymerase. Turns out you have to to a lot less cycles then with Phusion polymerase.

[pmiguel]

The slower running fragments after amplification is from 'hybrid' fragments, i.e. fragments that are created after primer becomes limiting and the DNA fragments anneal by the adaptors sequences at the ends but not in the 'unique' sequences in the middle. Think of a bubble structure.
[...]

Does not fit all the data I have seen. I have seen cases where higher order multimers are formed. That is, a 1-mer peak (expected length), 2-mer peak (a little less than double the expected length), 3-mer, etc.

This would fit with a daisy chain hypothesis but not with a bubble hypothesis.

--
Phillip

[ETHANol]

The slower running fragments after amplification is from 'hybrid' fragments, i.e. fragments that are created after primer becomes limiting and the DNA fragments anneal by the adaptors sequences at the ends but not in the 'unique' sequences in the middle. Think of a bubble structure. If you add fresh primer and do one more cycle of PCR (i.e. heat to 95 1 min, 60˚ C 10 30 sec 72˚C 30 sec), your fragments will return back to the expected size.

[pmiguel]

Bubble vs Daisy Chain vs Single Stranded vs PCR over-amplification

Reading this thread, and others on this topic, I get the impression that source of the extraneous high molecular weight products frequently visualized on bioanalyzer chips of TruSeq libraries is well understood.

However, I do not think this is the case!

Here are the hypotheses I as I understand them:

(1) "Bubble products". Until this morning I just assumed this was another way of describing "Daisy Chain" products (see below). But no. The name is the description -- two unrelated library molecules anneal to each other in the final cycle of PCR. Because their insert sequences are unrelated, only the adapter ends anneal, creating a big non-annealed "bubble" in the middle with annealed adapter "stems" on either end of the construct.

(2) "Daisy chain products". Two double stranded amplicons anneal at their adapter ends. Longer chains can form, for example a 3-mer, if both ends of an amplicon is annealed to other amplicons. Branching chains are also conceivable.

(3) "Single stranded products". This one is less clear to me. The post above, for example:

http://seqanswers.com/forums/showpost.php?p=45071&postcount=18

To get single stranded products from a PCR reactions, I think you need one primer to run out while the other persists. Then only one strand is being produced in each cycle and the amplification becomes linear, rather than exponential.

But single stranded products would electrophorese faster than double stranded products. The post above claims "they seem to appear higher on the Bioanalyzer". But I am pretty sure this is not the case.

[Note add 10/12/2011: The statement above is completely wrong! ssDNA tends to run larger (for long products, much larger) than dsDNA on Agilent Bioanalyzer chips. See:

http://seqanswers.com/forums/showthread.php?t=12852

for details.]

(4) "PCR overamplification products". Run a PCR reaction for too many cycles and you will get a high molecular weight smear. One possible source of this smear is that primers have been depleted and the products themselves anneal at their ends. Once annealed the polymerase will happily extend each single-stranded product into a chimeric dimer: adapter-insert1-adapter-insert2-adapter. Longer multimers could result as well.

My guess is that all 4 of these are at work in various situations. The length of the TruSeq adapters is probably the source of these issues. What about mitigation?
[Note added 12/25/2011 -- I don't recommend the following. As I indicated down-thread, it does not work. In fact it seems to increase the amount of multimers.]
It seems like (1) and (2) can be melted out if the right kinetics can be found. How about this:

TruSeq protocol call for PCR thermal cycling:

98°C for 30 seconds
10 cycles of:
— 98°C for 10 seconds #denaturation
— 60°C for 30 seconds #annealing
— 72°C for 30 seconds #extension
72°C for 5 minutes
Hold at 4°C

How about for the last cycle doing a very slow ramp from 98 oC to 60 oC? That way the full length products will have lots of time to find each other.

Or, skipping PCR altogether is great. Except for all the issues with getting an accurate titration for clustering...

--
Phillip

[aminoda]

I have also had the problem of the presence of the broad peak that is higher than the size-selected range from the Bioanalyzer analysis after PCR. As I have seen in some other thread, when I reduced the number of cycles from 15 to 10, and only using 2 ng DNA as starting material (measured with Qubit), it completely got rid of this problem. I was using all of whatever I had at that point before.

I've heard that the TruSeq adapter ligation is so efficient now, that too much DNA is going into the PCR reaction, and so the primers run out faster, creating ssDNA, which you cannot get rid of from running the gel and size selecting, since it runs at a similar speed as dsDNA (although they seem to appear higher on the Bioanalyzer). Trust me, I did exactly this, and did not make any difference...

[EMONGS]

BTS -

Rather than waiting we developed our own SPRI bead protocol. Gel cut samples captured with both Nimblegen and Illumina exome probe libraries are on HiSeq now. SPRI bead samples are in the queue.

John 1284...do you mind sharing your bead selection procedure?

Also, I have an issue with varying TruSeq DNA library sizes from 470 bp to 280 bp. I can't afford to spend time on troubleshooting right now so I was wondering if anyone has a similar problem and if so what is the remedy? Some of my initial samples were a little degraded. Will that make a difference or is it mainly the gel and excision?