mRNA library quantification BA/qbit/qPCR

[riehle]

Hi!

I want to quantify mRNA libraries produced with the TruSeq RNA Sample Preparation Kit from Illumina.

I have used Bioanalyzer, Qubit and qPCR (Kapa library quantification kit) and got, as expected, very different results. Especially between qPCR and the other two methods.
I have attached the bioanalyzer file of the libraries as well as word.doc that lists the molarities of the analyzed libraries.

Any comments from experienced lib prep people? Which meassurement should I trust? Are Bioanalyzer and Qbit usually underestimating the DNA concentration? Is this an expected result?

Any comments are welcome!

[HESmith]

This topic has been covered multiple times; please search the forum before posting a question.

[Christopher Odom]

I want to quantify mRNA libraries produced with the TruSeq RNA Sample Preparation Kit from Illumina.

I have used Bioanalyzer, Qubit and qPCR (Kapa library quantification kit) and got, as expected, very different results. Especially between qPCR and the other two methods.
I have attached the bioanalyzer file of the libraries as well as word.doc that lists the molarities of the analyzed libraries.

Any comments from experienced lib prep people? Which meassurement should I trust? Are Bioanalyzer and Qbit usually underestimating the DNA concentration? Is this an expected result?

Hi Riehle,

That is a great question. Many of our customers have reported a poor correlation between the concentrations of NGS libraries calculated with qPCR vs spectrophotometric methods (e.g. PicoGreen or Nanodrop) and electrophoretic methods (e.g using the LabChip GX or BioAnalyzer). The data set you provided is similar to what we've seen from other customers.

The conventional wisdom is that library samples would return a lower concentration when measured by qPCR, as opposed to a spectrophotometric method like PicoGreen, as qPCR "counts" only those molecules that have both adapters in the correct configuration, whereas spectrophotometric methods "count" all DNA molecules, irrespective of their adapter configuration. This is mostly true when libraries are quantified by Nanodrop, but the situation gets a bit more complicated when a PicoGreen assay is used.

As I’m sure you are aware, PicoGreen only effectively quantifies double-stranded DNA. During library amplification (unlike "normal" PCR from "biological" templates), the template is comprised of very high numbers of unique, short molecules. As a result, primers are the limiting component in the reaction. After a few cycles, there is no more free primer to anneal to the denatured template, and many molecules remain single-stranded at the end of the extension step. Since library molecules are all unique, there is statistically very little chance that a denatured molecule will find its complement again. The single-stranded library molecules that accumulate in this way now start to anneal to one another with their common and complementary adapter sequences, forming so called "daisy-chained" molecules. Depending on the library (template) and primer concentration at the start of the library amplification reaction, and the number of cycles of amplification performed, these "daisy-chained" molecules have different consequences for library quantification.

If the "daisy-chaining" starts happening early on in the amplification reaction (i.e. as a result of a high template and low primer concentration), it can lead to the generation of longer, chimeric molecules that are not quantified accurately by qPCR, PicoGreen or BioAnalyzer -- but these molecules do not sequence very effectively either. In most cases though, the "daisy-chained" molecules are bona fide library molecules that are just temporarily annealed to one another to form longer concatemers. The individual molecules will denature effectively during the initial denaturation step of a qPCR run, and during denaturation prior to flow cell amplification -- i.e. these annealed library fragments are still quantified reliably by qPCR and sequence perfectly well. PicoGreen and BioAnalyzer assays do not employ a DNA denaturation step. The "daisy-chained" molecules (which contain long stretches of single-stranded sequence) are therefore under-quantified with the PicoGreen assay, and are known not to migrate through gel matrices in a fashion similar to perfectly complementary dsDNA molecules (please see the attached example).

In the case of the data that you have attached, some of your samples appear to have the peak on the right-hand side of library peak which is likely the result of over-amplification. In these situations, the KAPA Library Quantification Kit will reliably quantify the “daisy-chained” species, while Picogreen will not. In addition, I think that you will find that if you were to run your samples on an Agilent High Sensitivity DNA Chip, the over-amplified product would appear more pronounced than you see on the Agilent DNA1000 Chip.

With respect to your question as to which data to trust, I would like to encourage you to send an annotated Excel file of your qPCR data, including average fragment lengths of your libraries to- [email protected]. That way, we can look at the data and get in touch with you to discuss the QC metrics to evaluate in order to determine whether library concentrations determined by qPCR are likely to be reliable.

Christopher Odom
KAPA Biosystems Technical Support

[pmiguel]

This topic has been covered multiple times; please search the forum before posting a question.

Actually, this is an issue we, and I think others, perpetually are struggling with.

My impression from reading your posts is that your have a great deal of expertise in this field. Of course you are under no obligation whatsoever to solve other people's problems. But, for myself, I would ask that you not stifle discussions of what -- it seems to me -- is a Achille's heel of Illumina library construction/sequencing.

--
Phillip

[HESmith]

Hi Phillip,

I'm not trying to stifle discussion, and I'm happy to share any expertise (however limited) I have to offer. However, this topic HAS been covered multiple times (and particularly well by you), and my intention was to direct the questioner to those threads. It's simply good web etiquette to search first, and the questions indicated to me that the questioner had not.

-Harold

[pmiguel]

Hi Harold,

I admit it has been covered multiple times. And maybe I am wrong to think that one more pass through the quagmire might show the right way.

But I keep seeing additional depths to the topic -- nearly every time we make another pass. For example, Chris Odom ventures into some territory regarding qPCR of these libraries that points to there being issues even with qPCR quantification of libraries.

As far as etiquette, I see your point. But if someone actually takes the effort to post their traces and qPCR results, I feel like they are not just lazily dumping their problems on us because they have taken the time to cogently format the question.

I may be a few steps ahead of them, but we are seeing so much variability that this whole area seems more like a "gaping wound" than a "problem solved". But maybe most of our travails are of the "Doc it hurts when I hold my arm like this" form.

That is, I was advised in another thread the no one "in their right mind" would multiplex deeper than 6 samples. Maybe they have a point. But I just see all the pieces there on the table and it seems like they should all fit together. There are just these strange inconsistencies that seem to plague our attempts to titrate libraries. For instance, I think, but can't yet prove that:

(1) Typically our libraries have non-dissolved, or even reversibly dissolved library-containing particulates in them such that we can't get our normal SYBR green qPCR methodology to approach giving consistent results without doing hard spins prior to every time we thaw them. (Dust, ampure beads, precipitated DNA?)
(2) Having too wide a size distribution in a library makes it well-nigh impossible to quantitate except by actually clustering the thing. (Your comments on this matter -- specifically issues with amplification efficiency were very much appreciated, by the way.)
(3) TruSeq libraries typically have some percentage of "hitch-hiking" primer/adapter-dimer single-strands annealed to full length molecules.
(4) A little EtBr in your library (from, for example, a PippenPrep size selection) can substantially throw off qPCR.

And I don't think this is the end of the weird stuff that is embedded in these protocols we are using.

Anyway, it may well be that endless discussion of these topics and the even more commonly discussed ones (bubble products, etc.) amount to little more than hand-wringing. And, sure, if we bring the signal/noise ratio of these forums too low, we will probably be driving away some who otherwise might contribute usefully.

So maybe for "borderline" posts of this sort, we just let slide but ignore or chastise anything that falls below this level of question?

--
Phillip

[pmiguel]

Hi Riehle,

That is a great question. Many of our customers have reported a poor correlation between the concentrations of NGS libraries calculated with qPCR vs spectrophotometric methods (e.g. PicoGreen or Nanodrop) and electrophoretic methods (e.g using the LabChip GX or BioAnalyzer). The data set you provided is similar to what we've seen from other customers.

[...]

Christopher Odom
KAPA Biosystems Technical Support

Hi Christopher,
Thanks for your post. I had a brief comment and a related question.

First, I would speculate that primer/adapter dimers are a major contributor to the formation of various DNA heteromultiplexes. Since primer/adapter dimers should be able to anneal to full library molecules, they could "hitch hike" along with the main peak--partially hiding even from gel cut size selection. Also you can imagine they might also serve as "splints" in various multimers that might form.

Second, I was wondering the extent to which KAPA qPCR can quantitate any given library in the presence of different amplification efficiencies library-to-library? That is, I presume some libraries just are more effectively amplified (closer to 100% efficiency) than others.

--
Phillip

[Christopher Odom]

First, I would speculate that primer/adapter dimers are a major contributor to the formation of various DNA heteromultiplexes. Since primer/adapter dimers should be able to anneal to full library molecules, they could "hitch hike" along with the main peak--partially hiding even from gel cut size selection. Also you can imagine they might also serve as "splints" in various multimers that might form.

Hi Philip,

There seems to be a possibility that the primers or adapter dimers do have some sort of a compounding effect on the formation of the heteromultiplexes, but we are still not certain exactly how the adapters interact. See the attached Bioanalyzer image where a single 470bp fragment was ligated to the adapters and three distinct peaks are seen, none of which corresponds to the unligated fragment. On the attached image we are likely seeing a species with adapters ligated to one end of the fragment, another with adapters ligated to both ends and then a third species of which the actual configuration is not known.

However, the fact that we see the multimeric complexes increasing as the number of amplification cycles increases supports the theory that the primers are being depleted and the single stranded molecules are annealing imperfectly to form the "daisy chain" molecules.

Second, I was wondering the extent to which KAPA qPCR can quantitate any given library in the presence of different amplification efficiencies library-to-library? That is, I presume some libraries just are more effectively amplified (closer to 100% efficiency) than others.

The qPCR Master Mix supplied in the KAPA Library Quantification Kit contains an engineered enzyme, that was specifically developed for high-efficiency SYBR Green I-based qPCR. Unlike wild-type DNA polymerases, this enzyme is capable of amplifying DNA fragments of diverse lengths and GC content with similar efficiency and low bias. Our kit is therefore designed to reliably "count" all the molecules in a complex DNA population.

Having said this, it has been widely reported that the correlation between library concentration and cluster density is not identical for all library types. In other words, if you quantify DNA libraries, small RNA libraries and CHiP-Seq libraries with our kit, and dilute them all to 10 pM, you will most likely not obtain the exactly same cluster density for each library type. It is not clear whether this is attributable to variation in reaction efficiency during qPCR due to library type, variation in cluster amplification due to library type or a combination of the two. Whilst we can guide you through the process of determining whether the results from a qPCR quantification run is reliable, it is up to every facility to establish the correlation between library concentration (determined by qPCR or any other quantification method) and cluster density -- for different library types and sequencing instruments. The best way to establish this correlation is to look at historical data (i.e. calculated library concentration vs actual cluster density) for as many samples as possible, grouped by by library type. From this type of analysis, you should be able to compensate for library type when planning how to dilute samples for cluster amplification.


Thanks again for your inquiry! For urgent assistance, please don’t hesitate to contact us directly at [email protected]

Christopher Odom
KAPA Biosystems Technical Support

[weigrc]

Dear Odom,

Since we are going to process ChIP-Seq libraries, can you explain more about "dilute them all to 10 pM"??!

And you said "likely not obtain the exactly same cluster ..."; so, we have experiences on sequencing of genomic DNA libraries and cDNA libraries, should we expect sequencing of ChIP-Seq libraries have higher or lower cluster density???


Thanks,
Wei

[Christopher Odom]

Since we are going to process ChIP-Seq libraries, can you explain more about "dilute them all to 10 pM"??!

Hi Wei,
The 10 pM concentration I mentioned in the last post was just an example. I suggest seeking out Illumina’s guidance with respect to the optimal input concentration for cluster amplification for different library types. I suspect that there is a target range, but that everyone ends up having to do some tweaking and optimization to obtain the best results for their specific samples and sequencing instruments.

And you said "likely not obtain the exactly same cluster ..."; so, we have experiences on sequencing of genomic DNA libraries and cDNA libraries, should we expect sequencing of ChIP-Seq libraries have higher or lower cluster density???

Based on customer feedback, and reports I have read in forums such as these, it is clear that the correlation between sample concentration (going into cluster amplification) and cluster density varies -- by library type, operator, instrument, etc. Unfortunately, I cannot offer specific recommendations on how to best dilute your ChIP-Seq libraries to obtain optimal cluster densities. My guess is that many members of this forum have struggled with the same question, so hopefully someone will be prepared to share some of their experiences.

I trust this is helpful. For urgent assistance, please contact us directly at [email protected]. Thank you!

Christopher Odom
KAPA Biosystems Technical Support