SMART-seq2 and Nextera XT DNA sample preparation kit

[wacguy]

Hi,

I was preparing 3 libraries for RNA-seq starting with 0.03, 0.3 and 3ng total RNA using the modified protocol for SMARTer ultra low (Picelli et al, 2014). I started the Tagmentaion/amplification step using the Nextera XT DNA sample prep kit with 1ng in each library. It looks to me very promising considering the low input but I think that there is still a place for improvement (see attached file). My impression, and from what I read, is that my DNA input for the Nextra protocol (1ng) is too high. I'm planning to try 3, 4 dilutions to see if I can get a smaller average fragment size and more "narrow" distribution but'd be happy for ideas and suggestions.

Thanks,
Guy

[Simone78]

you´re right, the tagmentation input is too high. I would suggest decreasing to 100-500 pg. Usually, when starting from 100 pg I do 12 cycles PCR while with 500 pg I do 10 cycles. You´ll get plenty of DNA anyway. I don´t know how your DNA looks like after pre-amplification, but your final library looks really good and I would expect nice results after sequencing.
Just one final observation. We noticed that the Nextera XT kit tends to give libraries with longer average size compared to the Nextera kit (the other kit, for input up to 50 ng DNA). This is probably due to the buffer, since Illumina bought Epicentre Tech. (which was making the original Nextera kit) and they kept the 2-buffers systems (HMW and LMW) they had. Therefore, you´ll never get a sharp peak when using the XT kit but rather a wide distribution. The goal is to keep the curve below 1000 bp, the limit for an efficient binding of the fragments to the flow cell.
Good luck!

[wacguy]

Thanks a lot for the detailed reply

I appreciate it!

[lbcbci]

you´re right, the tagmentation input is too high. I would suggest decreasing to 100-500 pg. Usually, when starting from 100 pg I do 12 cycles PCR while with 500 pg I do 10 cycles. You´ll get plenty of DNA anyway. I don´t know how your DNA looks like after pre-amplification, but your final library looks really good and I would expect nice results after sequencing.
Just one final observation. We noticed that the Nextera XT kit tends to give libraries with longer average size compared to the Nextera kit (the other kit, for input up to 50 ng DNA). This is probably due to the buffer, since Illumina bought Epicentre Tech. (which was making the original Nextera kit) and they kept the 2-buffers systems (HMW and LMW) they had. Therefore, you´ll never get a sharp peak when using the XT kit but rather a wide distribution. The goal is to keep the curve below 1000 bp, the limit for an efficient binding of the fragments to the flow cell.
Good luck!

Do you mean they are using HMW for Nextera kit and LMW for XT kit?
Thanks,

LL

[Simone78]

I´m not sure, but I don´t think Illumina has 2 different enzymes in the Nextera and Nextera XT kits. I think the the difference is the buffer. The profile we get with the 2 kits look very similar to the Epicentre buffers (LMW = Nextera; HMW = Nextera XT). And, again, based on my experience, the buffer are almost the same. It´s just a matter of having some additives in one and not in the other.

[Babooncanfly]

18s peak?

Hi op I'm wondering if you looked at cDNA size distribution by bioanalyzer, and saw any spike at 1.85kb? We saw this consistently, which is absent from Smarts-seq 1 protocol. Thanks in advance.

[Simone78]

Hi op I'm wondering if you looked at cDNA size distribution by bioanalyzer, and saw any spike at 1.85kb? We saw this consistently, which is absent from Smarts-seq 1 protocol. Thanks in advance.

We observed such a peak when using human cells but not cells from other organisms. Unfortunately, this is an artefact! In our case the peak comes from a mitochondrial transcript called "humanin", which has also several homologous copies on the genome. Humanins (there is actually a whole family) are involved in the stress response (among other things) and our hypothesis is that we observe it in the final library because the cell was stressed after all the manipulations it goes through before picking. Interestingly, we don´t always observe the peak even within the same cell type. At the moment we don´t have an explanation for it. Looking at the humanin sequence there is no clear complementarity with any of the oligos used in Smart-seq2 that would explain this preferential amplification (it means that there must be many copies of the transcript rather than a hotspot with some repetead seq that get picked up during the RT). We described a similar phenomenon in the Nat Methods paper and we observed that it is common between Smart-seq2 and the Clontech kit (see Suppl Fig 10 and main text). However, as you also pointed out, the 1.8 kb peak is not observed when using the Clontech kit.
Best regards,
Simone

[Babooncanfly]

We observed such a peak when using human cells but not cells from other organisms. Unfortunately, this is an artefact! In our case the peak comes from a mitochondrial transcript called "humanin", which has also several homologous copies on the genome. Humanins (there is actually a whole family) are involved in the stress response (among other things) and our hypothesis is that we observe it in the final library because the cell was stressed after all the manipulations it goes through before picking. Interestingly, we don´t always observe the peak even within the same cell type. At the moment we don´t have an explanation for it. Looking at the humanin sequence there is no clear complementarity with any of the oligos used in Smart-seq2 that would explain this preferential amplification (it means that there must be many copies of the transcript rather than a hotspot with some repetead seq that get picked up during the RT). We described a similar phenomenon in the Nat Methods paper and we observed that it is common between Smart-seq2 and the Clontech kit (see Suppl Fig 10 and main text). However, as you also pointed out, the 1.8 kb peak is not observed when using the Clontech kit.
Best regards,
Simone

Thanks Simone. I posted a similar message in another thread but thanks for answering here. What do you recommend that we do, just exclude humanin from future analysis? You still stand by the rest of the protocol being superior to clontech kit right?

[Simone78]

What we did (since we observed this peak long ago...) was to exclude the humanin reads from the analysis. And yes, Smart-seq2 is the only affordable alternative to the Clontech kit and the one we are using.
Please note that even the kit (as I already mentioned elsewhere, see Suppl Fig 10 in our Nat Methods paper) is generating libraries that have some weird "hotspots", that is part of exons (not even full exons or full genes) of some genes are covered by thousands and thousands of reads. This is clearly an artifact...and both Smart-seq 1 and 2 are affected!

[Babooncanfly]

Awesome thanks Simone!

[candida utilis]

SMART Seq V2 and Nextera XT

We aim to compare differential mRNA expression between 2 different cell types. We sorted 2 populations of cells by FACS, extract RNA by column based method, and SMARTSEQ v2 for pre-amplificaiton of cDNA.

Between those cell types, A has very low cell number (150-500 cells). In contrast, B has higher cell number (50,000 cells) but its RNA amount is less. So, I only do RNA quantification for B, because RNA amount in A is too low to be detected by Ribogreen.

For Pre amplification of cDNA, I have some questions:

(1) If the starting RNA amount used for SMARTSeq V2 is different between samples, will it affect the RNA Seq results, if we aim to compare the differential mRNA expression between those samples?

(2) For the pre-amplification of cDNA, will the number of PCR cycles for A and B need to be the same, if I would like to compare the differential mRNA expression between those cell types? If sample A need 20 cycles to get sufficient cDNA, while sample B only need 15 cycles, what should I do?

Thank you