Using RNA 6000 pico chips for final library QC

pmiguel · 07-05-2011, 12:49 PM

TruSeq libraries are frequently assayed on Bioanalyzer (Agilent 2100) DNA chips. However the "PCR enrichment" step in the standard TruSeq protocol frequently produces an extra band with a higher apparent molecular weight that confounds straightforward analysis of the range of amplicon sizes in a library. The nature of this extra band is uncertain. (See link for my speculation on this matter.)

Perhaps running strand denatured libraries on an RNA (single-stranded) Bioanalyzer chip is a better assay of library quality.

Here is a test we did (please skip the middle panel if it is confusing -- more about it later) :

The undenatured amplicons are bimodal in apparent molecular weight on both the DNA High Sensitivity chip and the RNA 6000 pico chip. However the undenatured size distribution is of little interest because the library will be alkali denatured prior to cluster PCR. Denaturation with heat, 95 oC for 2 minutes followed by "snap chill" on ice, (shown above) or dilution of 1 volume of library into 9 volumes of formamide with or without heat treatment (not shown) results in a single, presumably single-stranded, peak when run on the RNA chip.

Note that the center panel, showing the migrations and apparent MW of the non-denatured library on an RNA pico chip, does not accurately depict the length/MW of these double stranded DNAs unless they are strand-denatured. Please see this thread dsDNA migration on pico RNA chip for some strong evidence.

Note that there is a small amount of "adapter dimer" visible only in the pico RNA chip chromatogram. I think these are more-or-less standard results from a library made from RNA using the default RNA TruSeq kit/protocol. The TruSeq adapters are 63 and 58 nucleotides, for a total of 121 nucleotides. So the average "insert size" is about 150 bp for this library.

Decreasing the number of cycles of "enrichment PCR" is a valid method of decreasing the amount of the apparently higher MW band seen on DNA bioanalyzer chips. But the ability to see adapter dimer molecules might make it worth switching to an RNA (single stranded) chip instead. Something to think about anyway.
--
Phillip

Chipper · 07-05-2011, 02:45 PM

Nice experiment, we have had similar problems with ChIP samples where we have multipe peaks in the Bioanalyser after size-selection (2% agarose). I thought we had a problem with adaptor ligation and/or A-tailing since sizes correspond to adaptors + multiple inserts, but after reading your comments it is clear that the high mw bands are from annealing of fragments with different inserts. In some cases adaptor dimers will bind a single stranded fragment, which is why it is not visible in the hsDNA chip.

Adaptor dimers are from blunt-ended ligation without the A/T overhang, do you think it is caused by the oligosynthesis (if a small fraction of oligos are lacking the last base) or from exonuclease activity of the ligase (we used the NEB quick ligase). Perhaps using a normal ligase reaction would help since the quick ligase buffer supposedly enhances blunt-end liagtion?

pmiguel · 07-06-2011, 08:24 AM

It could also be exonuclease activity left over from an earlier end-polishing step. Do you know if your "T" overhang has an exonuclease-resistant linkage (phosphorothioate)?

--
Phillip

pmiguel · 07-06-2011, 11:23 AM

Quote:

Originally Posted by pmiguel

It could also be exonuclease activity left over from an earlier end-polishing step. Do you know if your "T" overhang has an exonuclease-resistant linkage (phosphorothioate)?

Oops. If the polishing enzyme used was T4 polymerase then phosphorothioate linkages apparently have no effect on its 3'->5' exonuclease activity. See:

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC320039/

But it might help with residual nuclease activity in your ligase...

--
Phillip

Chipper · 07-06-2011, 11:58 AM

We do column purification after polishing (epicentre kit) but perhaps the enzymes are not compleely removed. We will try to reduce the amount of adaptors, and do some additional cycles with excess primers for samples already made to see if we can get rid of the dimers.

pmiguel · 07-08-2011, 08:40 AM

Details below, but take a look at the same PCR enriched library from a TruSeq RNA sample and note the assay of the size difference between the DNA and the RNA chips:

Different library this time. Less fragmentation done on the input RNA. (4 minutes instead of 8) and fewer cycles of enrichment PCR (10 instead of 15).

To a first approximation, the nano and pico RNA chips give the same results (samples denatured 2 minutes 95 oC and "snap" cooled on ice prior to loading). But the DNA (high sensitivity) chip with no sample denaturation gives a very different lower size--showing almost nothing below 200 bp.

As to whether these lower molecular weight fragments are ssDNA "trapped" in higher molecular weight associations prior to heat denaturation or just ssDNA that does not register well on a DNA chip designed for dsDNA, I don't know.

--
Phillip

asaleh · 07-11-2011, 07:33 AM

Interesting Idea! I am curious how do the library concentration values on the different chips match up?

pmiguel · 07-11-2011, 12:19 PM

Quote:

Originally Posted by asaleh

Interesting Idea! I am curious how do the library concentration values on the different chips match up?

10, 21 and 16 ng/ul for DNA high sensitivity, RNA nano and RNA pico, respectively. According to the chips and after backing-out the dilutions used on 2 of them. My tendency would be to attribute this to the DNA chip not having much sensitivity in detecting ssDNA. My speculation, anyway.

--
Phillip

asaleh · 07-12-2011, 06:07 AM

I think those concentrations are more than close enough that qPCR could sort them out

pmiguel · 07-12-2011, 06:53 AM

Maybe. But size distribution given for the DNA and RNA chips is different for the same sample. So that is another factor.

Even if you ignore that the RNA chip shows detectable amounts of very short or even zero insert length amplicons where the DNA chip shows nothing below 200 bp, your calculations might be thrown off.

My guess is that the RNA distribution is probably closer to reality. That means there are small amplicons trapped in higher molecular weight associations in the dsDNA amplicons. That also means that size selections (Ampure) are not working as well as they might.

--
Phillip

Heisman · 07-12-2011, 10:48 AM

pmiguel, do you have next gen data sets where you also have done bioanalyzer traces of your libraries? You could compare the insert size distributions between the two data sets and probably figure out some interesting things. For example, if you see <200 bp fragments with the sequencing and you don't see them on the bioanalyzer chip, that implies your theory is correct about them traveling with larger fragments.

pmiguel · 07-12-2011, 11:55 AM

Not exactly. These are our first Illumina RNA libraries. (We have always used 454 or SOLiD in the past).

Might have something for you in a month or so...

--
Phillip

jlove · 07-28-2011, 08:50 AM

We've seen this double peak with PCR overcycling, and done the experiment where we took a little bit out and did 2 more cycles of PCR with it. The larger peak will always disappear, because it's chimera. We ultimately decided that when the two peaks arise, they are almost always 1X and 2X in sizing, and we figured that the second peak must be chimera of some form and should be ignored. (It won't affect sequencing products since there is a denaturation step.) We've been ignoring the larger peak when determining average size and using the smaller peak when using qPCR for concentration analysis.

pmiguel · 07-28-2011, 09:28 AM

Quote:

Originally Posted by jlove

We've seen this double peak with PCR overcycling, and done the experiment where we took a little bit out and did 2 more cycles of PCR with it. The larger peak will always disappear, because it's chimera. We ultimately decided that when the two peaks arise, they are almost always 1X and 2X in sizing, and we figured that the second peak must be chimera of some form and should be ignored. (It won't affect sequencing products since there is a denaturation step.) We've been ignoring the larger peak when determining average size and using the smaller peak when using qPCR for concentration analysis.

Sure, but you see that 126 nt peak on the pico chip above, right? I think that may be a player in forming the multimers you mention above.

No big deal when just a little is present, but what if you end up with lots of it? Then your qPCR result will be off and you might over shoot your target cluster density. (I mean, because the average amplicon size you use to calculate molarity, will not take into account the small amplicons.) If you use SYBR green qPCR, that is. I guess if you used actual Real Time PCR, then you would be okay.

--
Phillip

lilin001 · 03-04-2012, 08:38 PM

Quote:

Originally Posted by jlove

We've seen this double peak with PCR overcycling, and done the experiment where we took a little bit out and did 2 more cycles of PCR with it. The larger peak will always disappear, because it's chimera. We ultimately decided that when the two peaks arise, they are almost always 1X and 2X in sizing, and we figured that the second peak must be chimera of some form and should be ignored. (It won't affect sequencing products since there is a denaturation step.) We've been ignoring the larger peak when determining average size and using the smaller peak when using qPCR for concentration analysis.

Hi jlove,
My RNA library had two peaks, a ~600bp peak and a ~270bp peak. I am very excited to know that this doesn't affect sequencing. Could you please tell how many reads you got from your library? And what kit are using for qPCR? Thanks!

jlove · 03-05-2012, 07:42 AM

Hi lilin001,
I don't know about the number of reads (it varies from user to user how much we load and how many clusters we get), but when we do our calculations based on the size of the smaller "real" peak and the results of the qPCR (which is KAPA for Illumina library quantification by the way), we tend to yield a predictable number of clusters time and again. We don't get more adapter reads than usual with these samples, nor do we get more noise.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Typical Quantity of Primer Sequences In Final Library	Dario1984	Sample Prep / Library Generation	1	03-05-2012 07:44 AM
dsDNA migrates crazy fast on pico RNA chip	pmiguel	Sample Prep / Library Generation	10	10-31-2011 08:29 AM
Truseq Small RNA kit - losing library in the final cleanup?	Tali7	Sample Prep / Library Generation	3	10-12-2011 04:26 AM
Pico Green vs Real Time PCR	crndy52	454 Pyrosequencing	5	02-09-2011 01:18 AM
RNA-Seq: Comparison between NuGEN's WT-Ovation Pico and One-Direct Amplification Syst	Newsbot!	Literature Watch	2	10-04-2010 09:36 AM

07-05-2011, 12:49 PM	#1
pmiguel Senior Member Location: Purdue University, West Lafayette, Indiana Join Date: Aug 2008 Posts: 2,317	Using RNA 6000 pico chips for final library QC TruSeq libraries are frequently assayed on Bioanalyzer (Agilent 2100) DNA chips. However the "PCR enrichment" step in the standard TruSeq protocol frequently produces an extra band with a higher apparent molecular weight that confounds straightforward analysis of the range of amplicon sizes in a library. The nature of this extra band is uncertain. (See link for my speculation on this matter.) Perhaps running strand denatured libraries on an RNA (single-stranded) Bioanalyzer chip is a better assay of library quality. Here is a test we did (please skip the middle panel if it is confusing -- more about it later) : The undenatured amplicons are bimodal in apparent molecular weight on both the DNA High Sensitivity chip and the RNA 6000 pico chip. However the undenatured size distribution is of little interest because the library will be alkali denatured prior to cluster PCR. Denaturation with heat, 95 oC for 2 minutes followed by "snap chill" on ice, (shown above) or dilution of 1 volume of library into 9 volumes of formamide with or without heat treatment (not shown) results in a single, presumably single-stranded, peak when run on the RNA chip. Note that the center panel, showing the migrations and apparent MW of the non-denatured library on an RNA pico chip, does not accurately depict the length/MW of these double stranded DNAs unless they are strand-denatured. Please see this thread dsDNA migration on pico RNA chip for some strong evidence. Note that there is a small amount of "adapter dimer" visible only in the pico RNA chip chromatogram. I think these are more-or-less standard results from a library made from RNA using the default RNA TruSeq kit/protocol. The TruSeq adapters are 63 and 58 nucleotides, for a total of 121 nucleotides. So the average "insert size" is about 150 bp for this library. Decreasing the number of cycles of "enrichment PCR" is a valid method of decreasing the amount of the apparently higher MW band seen on DNA bioanalyzer chips. But the ability to see adapter dimer molecules might make it worth switching to an RNA (single stranded) chip instead. Something to think about anyway. -- Phillip

07-08-2011, 08:40 AM	#6
pmiguel Senior Member Location: Purdue University, West Lafayette, Indiana Join Date: Aug 2008 Posts: 2,317	Agilent DNA high sensitivity vs. Nano vs. Pico Details below, but take a look at the same PCR enriched library from a TruSeq RNA sample and note the assay of the size difference between the DNA and the RNA chips: Different library this time. Less fragmentation done on the input RNA. (4 minutes instead of 8) and fewer cycles of enrichment PCR (10 instead of 15). To a first approximation, the nano and pico RNA chips give the same results (samples denatured 2 minutes 95 oC and "snap" cooled on ice prior to loading). But the DNA (high sensitivity) chip with no sample denaturation gives a very different lower size--showing almost nothing below 200 bp. As to whether these lower molecular weight fragments are ssDNA "trapped" in higher molecular weight associations prior to heat denaturation or just ssDNA that does not register well on a DNA chip designed for dsDNA, I don't know. -- Phillip

07-05-2011, 02:45 PM	#2
Chipper Senior Member Location: Sweden Join Date: Mar 2008 Posts: 324	Nice experiment, we have had similar problems with ChIP samples where we have multipe peaks in the Bioanalyser after size-selection (2% agarose). I thought we had a problem with adaptor ligation and/or A-tailing since sizes correspond to adaptors + multiple inserts, but after reading your comments it is clear that the high mw bands are from annealing of fragments with different inserts. In some cases adaptor dimers will bind a single stranded fragment, which is why it is not visible in the hsDNA chip. Adaptor dimers are from blunt-ended ligation without the A/T overhang, do you think it is caused by the oligosynthesis (if a small fraction of oligos are lacking the last base) or from exonuclease activity of the ligase (we used the NEB quick ligase). Perhaps using a normal ligase reaction would help since the quick ligase buffer supposedly enhances blunt-end liagtion?

07-06-2011, 08:24 AM	#3
pmiguel Senior Member Location: Purdue University, West Lafayette, Indiana Join Date: Aug 2008 Posts: 2,317	It could also be exonuclease activity left over from an earlier end-polishing step. Do you know if your "T" overhang has an exonuclease-resistant linkage (phosphorothioate)? -- Phillip

07-06-2011, 11:58 AM	#5
Chipper Senior Member Location: Sweden Join Date: Mar 2008 Posts: 324	We do column purification after polishing (epicentre kit) but perhaps the enzymes are not compleely removed. We will try to reduce the amount of adaptors, and do some additional cycles with excess primers for samples already made to see if we can get rid of the dimers.

07-11-2011, 07:33 AM	#7
asaleh Member Location: Maryland Join Date: Jun 2011 Posts: 16	Interesting Idea! I am curious how do the library concentration values on the different chips match up?

07-12-2011, 06:07 AM	#9
asaleh Member Location: Maryland Join Date: Jun 2011 Posts: 16	I think those concentrations are more than close enough that qPCR could sort them out

07-12-2011, 06:53 AM	#10
pmiguel Senior Member Location: Purdue University, West Lafayette, Indiana Join Date: Aug 2008 Posts: 2,317	Maybe. But size distribution given for the DNA and RNA chips is different for the same sample. So that is another factor. Even if you ignore that the RNA chip shows detectable amounts of very short or even zero insert length amplicons where the DNA chip shows nothing below 200 bp, your calculations might be thrown off. My guess is that the RNA distribution is probably closer to reality. That means there are small amplicons trapped in higher molecular weight associations in the dsDNA amplicons. That also means that size selections (Ampure) are not working as well as they might. -- Phillip

07-12-2011, 10:48 AM	#11
Heisman Senior Member Location: St. Louis Join Date: Dec 2010 Posts: 535	pmiguel, do you have next gen data sets where you also have done bioanalyzer traces of your libraries? You could compare the insert size distributions between the two data sets and probably figure out some interesting things. For example, if you see <200 bp fragments with the sequencing and you don't see them on the bioanalyzer chip, that implies your theory is correct about them traveling with larger fragments.

07-12-2011, 11:55 AM	#12
pmiguel Senior Member Location: Purdue University, West Lafayette, Indiana Join Date: Aug 2008 Posts: 2,317	Not exactly. These are our first Illumina RNA libraries. (We have always used 454 or SOLiD in the past). Might have something for you in a month or so... -- Phillip

07-28-2011, 08:50 AM	#13
jlove Member Location: Boston, MA Join Date: Apr 2011 Posts: 49	We've seen this double peak with PCR overcycling, and done the experiment where we took a little bit out and did 2 more cycles of PCR with it. The larger peak will always disappear, because it's chimera. We ultimately decided that when the two peaks arise, they are almost always 1X and 2X in sizing, and we figured that the second peak must be chimera of some form and should be ignored. (It won't affect sequencing products since there is a denaturation step.) We've been ignoring the larger peak when determining average size and using the smaller peak when using qPCR for concentration analysis.

03-05-2012, 07:42 AM	#16
jlove Member Location: Boston, MA Join Date: Apr 2011 Posts: 49	Hi lilin001, I don't know about the number of reads (it varies from user to user how much we load and how many clusters we get), but when we do our calculations based on the size of the smaller "real" peak and the results of the qPCR (which is KAPA for Illumina library quantification by the way), we tend to yield a predictable number of clusters time and again. We don't get more adapter reads than usual with these samples, nor do we get more noise.