[HeidelbergScience]

Why do you use T4 PNK instead of a regular phosphatase (e.g. SAP)? Do you need 5’ phosphates for some reason?

Also, why do you say sonication results in 50% 5’ phosphates, 50% 3’? The literature discussed in this thread says there's a "great preference" for 5’: http://seqanswers.com/forums/showthread.php?t=2759

Have you tried using sonicated DNA without phosphatase treatment?

T4 PNK removes cyclo-phosphate and phosphate from 3-OH of RNA and DNA, but in the presence of ATP is also phosphorylates 5-OH. For CATS you need only to free the 3-OH group. The method doesn't care about a 5'-phos, and therefore there is no need to add ATP together with with T4 PNK.

https://www.neb.com/products/m0201-t4-polynucleotide-kinase

After sonication at least some % of DNA fragments should have 3'-phophates, although its proportion may vary from condition to condition. One should repair the 3'-ends to be on a safe side.

[jwfoley]

Why do you use T4 PNK instead of a regular phosphatase (e.g. SAP)? Do you need 5’ phosphates for some reason?

Also, why do you say sonication results in 50% 5’ phosphates, 50% 3’? The literature discussed in this thread says there's a "great preference" for 5’: http://seqanswers.com/forums/showthread.php?t=2759

Have you tried using sonicated DNA without phosphatase treatment?

[HeidelbergScience]

Importantly, the CATS protocol described in the paper (and in this forum) is optimized and tested only for 1 pg/µL - 10 ng/µL (final concentration in polyA / polydA tailing reaction) of RNA and DNA amounts. When using higher inputs (>10 ng/µL), additional optimization might be necessary (e.g. ATP/dATP concentration, primers concentration, incubation times).

[HeidelbergScience]

Why the 3' V anchor instead of the more common VN, on your oligo(dT) RT primer?

The V primer has 4x higher efficient concentration as compared to VN, and it also decreases the probability of the formation of "empty" libraries.

[HeidelbergScience]

Hi EvilTwin,

Yes you did understand my question. Thanks for the response. Can CATS be used at equivalent quantities as the other kits? E.g. if you added 100ng into the reaction would it flood the system or create negative effects compared to other kits?

We never tested >10 ng of DNA or RNA per reaction. However, if you use 10 ng (300 bp DNA or RNA) - the product is already strongly visible after pre-amp cycle 8 (if using the optimized protocol described in this forum). We never used less than 8 PCR cycles for the pre-amp, because pre-amp is aimed to generate dsDNA libraries with the Illumina P5 and P7 adaptors from single-stranded first cDNA strand.

Importantly, the complexity of libraries generated from >10 ng nucleic acids will be unnecessary high. Thus, 10 ng of input should yield a library with the complexity (a number of different fragments) that already exceeds the capacity of a 200M-read HiSeq run. For example: already 10 ng of 300 bp DNA is approximately 30 000 000 000 different molecules. If we assume that the efficacy of TSO at those concentration is close to 80%, the libraries would contain a similar number of different reads.

[jwfoley]

Why the 3' V anchor instead of the more common VN, on your oligo(dT) RT primer?

[Nighthawkrao77]

Hi EvilTwin,

Yes you did understand my question. Thanks for the response. Can CATS be used at equivalent quantities as the other kits? E.g. if you added 100ng into the reaction would it flood the system or create negative effects compared to other kits?

[EvilTwin]

Is there evidence that you get very high RNA whole transcriptome coverage using this method compared to others? I know in the paper it claims that this comparison wasn't done because input/cost are also floating variables.

Hi Nighthawkrao,


which metrics are of predominant interest for you?


In terms of library complexity on whole transcriptome for 1 ng (rRNA-depleted) and below, CATS gives very good results that are comparable to using > 50 ng with other kits.



We have verified the expression levels of several highly as well as lowly expressed genes in our CATS runs with conventional RT-PCR and get a very good concordance.


A direct comparison to conventional kits (using the same amounts) is truly not easy because they do not work with the CATS low input ranges.


The exact coverage of the transcriptome/certain genes depend on the composition of the RNA sample (what type of RNA is predominant) origin (cell line/tissue/plasma etc, cell cycle/expression status of genes of interest, etc) and aside from the input is a function of fragment complexity captured during library generation (which is the job of CATS) and of how many reads you actually sequence.


Did I understand your question correctly?

[Nighthawkrao77]

Is there evidence that you get very high RNA whole transcriptome coverage using this method compared to others? I know in the paper it claims that this comparison wasn't done because input/cost are also floating variables.

[Simone78]

At the time of the preparation of our manuscript we did not found scientific reports describing the application of the DNA Nextera XT kit. Isn’t it quite new? The manual of DNA Nextera kit stated that DNA amount should be not less than 50 ng and the length >2000 bp. So, we suppose that the statement in the paper was correct, albeit already outdated.

I just randomly googled "Nextera XT + launch + date" and I got this:

https://www.genomeweb.com/sequencing/new-products-illuminas-nextera-xt-kits-agilents-haloplex-system-ion-torrent-kits

so, already in mid-2012 the Nextera XT was available/going to be available. in fact, when we first tried the XT kit and compared to our first version of home-made tn5 was mid-2013 for sure...but this is not so important.
The 50 ng and 2000 bp are meant as the "golden standard". If you, for example, tagment a library of 1000 bp (avg size) you´ll get usable data anyway but probably will have a poor coverage at the 5´end of the transcripts, especially the long ones. It depends what you are looking for with your experiments, of course.

Although we cited only one paper on Tn5, our statement that “The full capacity of the tagmentation technique for DNA library prep is yet to be tested and compared with other methods” is correct because none of the papers described application Tn5 for DNA-seq of low amounts of fragmented circulating DNA.

Importantly, Tn5-based method are apparently restricted to dsDNA of >300 bp (inferred form Nextera XT manual), and might not be efficient for very fragmented (<150 bp) circulating DNA. Again, we are not claiming that it is impossible; we just saying that “it has to be tested and compared”. Despite the fact that tagmentation may occur on short (<150 bp) dsDNA, the complexity of such libraries has to be demonstrated.

Yes, the Nextera or Nextera XT manuals can tell you whatever they want...but doesn´t necessarily means that it´s true! If you look, again, to Adey et al. (2010) you will see that even fragments down to 40 bp can be tagmented (what we also observed with our Tn5). Of course Adey et al. were not working with fragmented DNA but the fact that the tn5 cuts even very short fragments remains. Besides, the transposase is not a "classic" enzyme, so each dimer cuts and ligates DNA only once and then becomes inactive. What we observed is that the cutting efficiency can be increase with additives (PEG), increasing the amount of enzyme, decreasing the amount of DNA and/or extending the incubation time. As we also observed, if you have a cDNA library with a very short avg size after Smart-seq2 (derived from a very degraded sample) you will get a very beautiful peak after tagmentation+PCR...but when you´ll analyze the data it will look awful. Translation: the tn5, especially when added in excess, can cut virtually ANYTHING that is double-stranded (and the reaction is driven to completion). The complexity reflects the initial quality of your sample rather than the efficiency of the tagmentation reaction alone. Or, at least, this is what we concluded looking at the data.

Finally, the library preparation workflow of the published Tn5 methods for bisilfiteDNA-seq is significantly more labor and (at the moment) cost intensive as compared to CATS. However, we certainly agree that tagmentation is a very elegant and promising method, especially as compared to adaptors-ligation.

not really, just the oligo replacement step. Once you can make the Tn5 yourself it´s not big deal playing with the adaptors. About the costs you were right but it´s not true anymore if one can make his own tn5.

All in all, I really appreciate your comments and explanations! I´ll have them with me when I will read the paper again! I don´t have time to go through all your answers to my comments now but I think they addressed most of my doubts (some remain...). And, as I said in the beginning: the paper has some interesting ideas and interesting possible applications!

[Simone78]

Our statement is correct, because Ribozero-treated RNA and polyA-enriched RNA is not the same. The paper underlying the stranded RNA-seq kit from Clontech was cited in our paper as well [Langevin et al, 2013 RNA Biol].

It was just to give an idea of the amounts we were talking about. ASSUMING 10 pg TOT RNA/cell AND ASSUMING that 1-5% of the cell RNA is mRNA, then 1 ng mRNA corresponds APPROXIMATELY to 200 pg - 1 ng TOT RNA. So the inputs to start with the different protocols are in the same order of magnitude.

[Simone78]

Unlike in CATS, in SMART-seq the TS at the end of the RNA templates may not be a rate limiting step; since during SMART the most template switching events are likely to occur before the MMLV RT reaches the end of the mRNA (and starts to add dCs). Albeit only a speculation, but this could be the reason why you did not observe the increase after Mn2+ addition.

exactly! unfortunately just an UNSUPPORTED speculation...however we also tried to add the MnCl2 at different times after the RT was initiated. Result: the sooner the MnCl2 was added the lower the final yield up to a point that the yield was zero when MnCl2 was added in the beginning. Of course this might also be due to some oxidation/interference with some other component in the reaction.
Besides, if we had so much "premature" TS then we wouldn´t observe an almost even coverage of the transcript body or of the longest transcripts as we, in fact, do (it´s actually one of the major improvements over the Clontech kit...).

[Simone78]

However, to obtain Tn5 in-house one need to produce and purify it from mammalian cells, if we understood correctly? While this might be possible at some institutions, CATS would be a simpler and a cheaper way where this is not possible. Or, is there is already a cheap commercial provider of Tn5?

We made available the pTBX1-Tn5 plasmid to everybody and it can be purchased from Addgene for 65 USD. It can be easily produced in E.coli (C3013 strain). 1 litre of culture will make as much tn5 as you need for probably half a million single cell experiments.

[HeidelbergScience]

If you would use Ribozero + Smarter (the STRANDED kit, maybe I should have specified it. I thought it was clear from the context but if you are not familiar with the Clontech kits it might have been confusing, sorry) then what you claim at point 1 and 3 above is wrong. Of course, Smart-seq2 is unfortunately still limited to mRNA only (and some additional other bias and problems). About point 4: I didn´t count exactly the mins you need to do a plate with Smart-seq2 but I don´t think is so far way from your protocol in terms of hands-on time. Every step can be easily done on the Bravo and almost all the master mixes can be prepared even weeks in advance and thawed just before adding them to the sample.
About point 2: in our Nat Methods paper (tagmentation done with Nextera) the coverage at the 3´and 5´is analysed in Suppl Fig 3E, 8 and 9 (as well as in Figure 2 of the main text). You will notice that there is no drop at the 3´and also a good coverage with only a slight drop at the 5´. Explanation: at both ends we have some "extra template" for the transposase to cut, i.e. the long oligo dT and the TSO. It means that even if the first 10, 20 or 30 nucleotides are "lost" with the cutting the transcript is not affected. In our recent Genome Res paper we saw the same but we didn´t report it since the library was done again with Smart-seq2 and there was no difference in performance between Nextera and our transposase in all the things we looked at.

We know all Clontech kits of course, but we thought you meant SMARTer Universal Low Input RNA Kit and your-own SMART-seq2, as you wrote “10 ng + SMARTer (or SMART-seq2)”.

For SMARTer Stranded RNA-Seq Kit points 1, 3 (partly) and 4 definitely cannot be applied; however the most important question is - what is the complexity of the final library. In fact even conventional ligation-based methods may generate the library from 5 ng RNA or DNA (or even below). The question is – how many different fragments they would capture, and what would be the % of PCR duplicates.

SMARTER stranded kit is based on random N6 priming, and therefore: restricted to only long RNAs. It certainly will not capture short RNAs (miRNA, piRNA) and have troubles with RIP samples. Furthermore, random priming RT is inefficient on DNA templates, since DNA/DNA binding is much weaker than RNA/DNA.

The fundamental advantage of poly(A)/(dA) tailing is that even 1 molecule in the solution will be tailed (there is no limit of "efficient concentration"), while the use of long (30x) polydT reverse primer secures the efficient capture of highly diluted molecules for the RT.

Also, good to know that the SMART-seq2 has even representation over the mRNA. It remains the great option for single-cell mRNA sequencing indeed.

[Simone78]

One can use 10 ng of total RNA input, enrich it for mRNA via poly(dT) magnetic beads and fragment with Mg2+, yielding approx. 100-300 pg of 20-100 bp RNA in 40 µl eluate. The important thing here is to add 10 µg of glycogen as a co-precipitant during clean-up step (with miRNAeasy kit). In our hands, the whole procedure of sample preparation (before polyA-tailing) takes about 1 hour. Subsequently one could set-up a poly(A) reaction in 50 µl and, than concentrate the whole 50 µl (e.g. via Zymo columns or EtOH precipitation). Even 10 pg of fragmented RNA is already enough for CATS (the protocol posted on this forum) to generate high complexity libraries for mRNA-seq. There are also many other options to fragment RNA (including thermosensitive RNAses) w/o the need to subsequently purify and concentrate the sample.

However, if we speak about mRNA-seq from ultra-low (1-100) number of cells, then SMART-seq is probably the most convenient way (and Ribozero would not be even necessary). However, most researchers are working with much higher number of cells (e.g. growing on 96 - 24well plates), from where obtaining 100-1000 ng of RNA is not a problem. After ~30 min procedure of poly(A)-enrichment, one can get 3 – 30 ng of mRNA in 40 µl eluate, and run CATS even without a need to concentrate the polyadenylated product before RT. Also, unlike SMART-seq, CATS gives (1) strand-specific information about mRNA and (2) has even coverage along all mRNAs. While:

(1) SMART-based mRNA-seq is not strand specific.

(2) with SMART-seq 5’-proximal and 3’-proximal parts of mRNAs are likely to be significantly underrepresented due to the inevitable premature template switching and taqmentation bias. Please correct us if we are wrong.

(3) SMART-seq is limited only to mRNA-sequencing; while CATS allows any RNA-seq, including small (20-200nt) RNA, like RIP-samples, miRNA, piRNA etc, and also any DNA-seq. So it is a universal protocol.

(4) SMART-seq would actually require more efforts because “RT and template switch” there is used only to generate and pre-amplify long cDNAs from mRNA. The library preparation itself occurs afterwards via fragmentation/adaptors ligation or taqmentation and further pre-amplification + purification. It is much easier to do mRNA enrichment (30 min), fragmentation/cleanup (20 min) and CATS (4-5 hours total, 20 min hands-on time).

To summarize, if you have a few cells and require only mRNA-seq than SMART-seq is the probably the best option. However, one can still convert mRNA from a single-cell into cDNA using poly(dT) primers, and run CATS after genome-wide DNA pre-amplification till hundred picograms.

If you would use Ribozero + Smarter (the STRANDED kit, maybe I should have specified it. I thought it was clear from the context but if you are not familiar with the Clontech kits it might have been confusing, sorry) then what you claim at point 1 and 3 above is wrong. Of course, Smart-seq2 is unfortunately still limited to mRNA only (and some additional other bias and problems). About point 4: I didn´t count exactly the mins you need to do a plate with Smart-seq2 but I don´t think is so far way from your protocol in terms of hands-on time. Every step can be easily done on the Bravo and almost all the master mixes can be prepared even weeks in advance and thawed just before adding them to the sample.
About point 2: in our Nat Methods paper (tagmentation done with Nextera) the coverage at the 3´and 5´is analysed in Suppl Fig 3E, 8 and 9 (as well as in Figure 2 of the main text). You will notice that there is no drop at the 3´and also a good coverage with only a slight drop at the 5´. Explanation: at both ends we have some "extra template" for the transposase to cut, i.e. the long oligo dT and the TSO. It means that even if the first 10, 20 or 30 nucleotides are "lost" with the cutting the transcript is not affected. In our recent Genome Res paper we saw the same but we didn´t report it since the library was done again with Smart-seq2 and there was no difference in performance between Nextera and our transposase in all the things we looked at.