RNA-seq from single cells?

[SunPenguin]

Hey Simone,

DOI: 10.1038/ncomms11112

Is strand invasion a big issue in your hands? The paper that was cited before tested using 3 LNA contigs (+G+G+G), and I wonder if using the rGrG+G configuration more or less avoids that problem.

I am also looking around at these different template switching protocol. How important is it to clean the first strand cDNA after synthesis? Some protocols even try to chew away the TSOs with exonuclease (by including some rU's throughout the sequence).

For the bulk sample stuff, I've never cleaned the first strand cDNA, and everything seems to be okay. Is it something that's more important for single cells?

[SunPenguin]

Hey Serena,

I've been playing around with several different versions of the TSOs, but for the single cell stuff, I've been using basically the exact same sequence as published in the original Smart-seq2 paper, but with the 5'biosg addition. the blocked oligos more or less avoids the concatemers (though sometimes I still get a little bit).

I have been using poly(dT) with 3'VN though. Perhaps that's part of the issue.

From my experience though, if you're not getting results from single cell sort, but is consistently getting results from 10 cells or more, the more likely issue is the sorting. I'm sure you're already doing this, but you have to fastidiously calibrate your sorter for single cell sort (I personally use a tiny hole puncher to punch out small circles of PH papers to detect where exactly a droplet falls in the well. Extremely time consuming, but it's the only way I can ensure good sorting).

Also, snap freeze your samples after sort, if you don't already. At least, put it directly on a layer of dry ice asap after sorting.

Also switch to Maxima H- RT if you haven't already. We get better success rate (within samples of the same sorting session) with maxima RT than with SSII or with SmartScribe. 42 degrees is better than 50, for Maxima.

[Simone78]

Originally Posted by SunPenguin

Is strand invasion a big issue in your hands? The paper that was cited before tested using 3 LNA contigs (+G+G+G), and I wonder if using the rGrG+G configuration more or less avoids that problem.

I think the strand invasion is difficult to quantify but it is definitely a problem, even when using "only" one LNA-G.

Originally Posted by SunPenguin

I am also looking around at these different template switching protocol. How important is it to clean the first strand cDNA after synthesis? Some protocols even try to chew away the TSOs with exonuclease (by including some rU's throughout the sequence).

For the bulk sample stuff, I've never cleaned the first strand cDNA, and everything seems to be okay. Is it something that's more important for single cells?

I think doing the purification after 1st strand synthesis is not necessary, even in single cells. However, we observed sub-optimal performances of the KAPA Pol when trying to reduce the volume of the pre-ampl reaction too much. The reason is probably that the salt and additives from the RT are disturbing the PCR. If you dilute them by adding 1.5-2 volumes of PCR mix the reaction works very well. Increasing the final volume even more probably make things even better but it becomes very expensive when you need to process hundreds cells and we don´t do it.

Exonuclease I will digest only ssDNA in direction 3´-->5´, basically only the free oligos not bound to the cDNA. The Quarz-seq protocol uses ExoI after RT but they don´t do the strand switching so they have only the hybrid mRNA:cDNA at the end of the RT (and blunt ends). I tried to do it as well (only once) but the following PCR failed completely and I never repeated it again. I must have done something wrong because, in theory, the TSO is annealed to the cDNA and gets extended, so the 5´portion of the hybrid cDNA:mRNA should actually be dsDNA and therefore not accessible for the ExoI.

[skwek1]

Originally Posted by SunPenguin

Hey Serena,

I've been playing around with several different versions of the TSOs, but for the single cell stuff, I've been using basically the exact same sequence as published in the original Smart-seq2 paper, but with the 5'biosg addition. the blocked oligos more or less avoids the concatemers (though sometimes I still get a little bit).

I have been using poly(dT) with 3'VN though. Perhaps that's part of the issue.

From my experience though, if you're not getting results from single cell sort, but is consistently getting results from 10 cells or more, the more likely issue is the sorting. I'm sure you're already doing this, but you have to fastidiously calibrate your sorter for single cell sort (I personally use a tiny hole puncher to punch out small circles of PH papers to detect where exactly a droplet falls in the well. Extremely time consuming, but it's the only way I can ensure good sorting).

Also, snap freeze your samples after sort, if you don't already. At least, put it directly on a layer of dry ice asap after sorting.

Also switch to Maxima H- RT if you haven't already. We get better success rate (within samples of the same sorting session) with maxima RT than with SSII or with SmartScribe. 42 degrees is better than 50, for Maxima.

Hi SunPenguin,
Thank you for your response! I have been using oligodtVN, but Simone has suggested using just oligodtV without the N according to the CAT publication. What are your thoughts about this?

I calibrate my single cell sort with 100 cells onto a film on top of the plate. I can see it is not as good as you what you have suggested. Nevertheless there are application of cDNA from single cell with 21 cycles, just not a lot. After sorting, I vortex on a plate shaker, spin down and snap freeze on dry ice.

There are definitely less dimers now (but still some) when I biotinylated all the oligos. I'm not sure if concatamers are a problem for me as I see some jagged peaks and they are not constant like for concatamers.

Is this amount of cDNA enough to proceed for single cell? Please see attached bioanalyzer profile for 0 and 1 cell.

I will try 23 cycles as Simone as suggested, as well as rGrGrG without LNA, and also Maxima H RT as you have suggested.

Thank you!
Serena

[SunPenguin]

How much lysis/RT buffer do you sort into?

The poly(dT)VN vs without VN is also new to me. I can see how it can make some weird binding situation with the template switch oligo.

For what it's worth, Some people at the Broad think they both work the same.

[achamess]

Originally Posted by Simone78

I think the strand invasion is difficult to quantify but it is definitely a problem, even when using "only" one LNA-G.



I think doing the purification after 1st strand synthesis is not necessary, even in single cells. However, we observed sub-optimal performances of the KAPA Pol when trying to reduce the volume of the pre-ampl reaction too much. The reason is probably that the salt and additives from the RT are disturbing the PCR. If you dilute them by adding 1.5-2 volumes of PCR mix the reaction works very well. Increasing the final volume even more probably make things even better but it becomes very expensive when you need to process hundreds cells and we don´t do it.

Exonuclease I will digest only ssDNA in direction 3´-->5´, basically only the free oligos not bound to the cDNA. The Quarz-seq protocol uses ExoI after RT but they don´t do the strand switching so they have only the hybrid mRNA:cDNA at the end of the RT (and blunt ends). I tried to do it as well (only once) but the following PCR failed completely and I never repeated it again. I must have done something wrong because, in theory, the TSO is annealed to the cDNA and gets extended, so the 5´portion of the hybrid cDNA:mRNA should actually be dsDNA and therefore not accessible for the ExoI.

Hi. Thanks so much for all of the info you provide and for developing Smart-seq2.

I used Smart-seq2 the other day for amplifying mRNA from single neuronal nuclei, as well as bulk (100 nuclei). It worked great, the first time! The singles amplified very nicely. I used 21 cycles, as advised by Krishnaswami et al (2016) in their Nat. Protocol. I got plenty of cDNA, and nothing from the buffer that contained the nuclei (NTC).

I was really pleased. FYI, I'm using biotinylated TSO, ISPCR and oligoDT.

My question is about the maximum number of cells/nuclei I can use. From the Nat protocol, I see the recommendation to not dilute the lysis buffer any more than 0.5 ul. When I sorted, I was certainly below this. but if I wanted to increase the number of nuclei that I process in bulk (say, 1000+), I would be going up to about 5 ul of sorting buffer (PBS), which would mess up the concentrations for the downstream Smart-seq2 procedure.

So I was wondering if you have any guidance. Can I scale up? Or, alternatively, I could sort into more lysis buffer (say 5 ul lysis buffer), lyse the cells and homogenize their mRNA, and then take a fraction of that lysis solution (2 ul) for downstream prep. That way I would be getting a representation of more nuclei (say 1000) but with only a fraction of their input. Is that advisable?

Thanks so much for your help.

[skwek1]

Originally Posted by SunPenguin

How much lysis/RT buffer do you sort into?

The poly(dT)VN vs without VN is also new to me. I can see how it can make some weird binding situation with the template switch oligo.

For what it's worth, Some people at the Broad think they both work the same.

2ul as in the Smartseq2 protocol.

Thanks,
Serena

[SunPenguin]

And do you recover all of 2uL? I usually sort into 5 uL (better chance of hitting the buffer). By the time I freeze, thaw, and actually pipette in the volumes to do RT, I get about 2-3uL back (I pipette out the lysate from the plate, since I don't usually process a whole plate at a time, yet). Evaporation happens fast with such small volume, even during the wait time on the sorter. Or unless you're sorting into 384 well plates, then the format is a bit different.

[skwek1]

Originally Posted by SunPenguin

And do you recover all of 2uL? I usually sort into 5 uL (better chance of hitting the buffer). By the time I freeze, thaw, and actually pipette in the volumes to do RT, I get about 2-3uL back (I pipette out the lysate from the plate, since I don't usually process a whole plate at a time, yet). Evaporation happens fast with such small volume, even during the wait time on the sorter. Or unless you're sorting into 384 well plates, then the format is a bit different.

Hi SunPenguin,
I sort into 96 well PCR plates with 2 ul of lysis buffer containing TritonX and Rnase In (same % as in paper), froze down and thaw as needed. Upon thawing, I vortex on a shaker and spin down before removing the microamp clear adhesive film. I add 2ul of premixed oligodtVN and dNTP, and do all subsequent reactions in the same plate.
Thanks for working with me on this!
Serena

[naveenrajkumar]

[QUOTE=Simone78;191832]

I think doing the purification after 1st strand synthesis is not necessary, even in single cells. However, we observed sub-optimal performances of the KAPA Pol when trying to reduce the volume of the pre-ampl reaction too much. The reason is probably that the salt and additives from the RT are disturbing the PCR. If you dilute them by adding 1.5-2 volumes of PCR mix the reaction works very well. Increasing the final volume even more probably make things even better but it becomes very expensive when you need to process hundreds cells and we don´t do it.


Hi Simone,
what is the lowest total volume of preamp reaction have you successfully performed using KAPA? or when do you notice sub-optimal performance of KAPA. what was the ratio of RT in KAPA mix. I notice in your paper you use 20% cDNA in PCR volume (10ul in 50ul). Have you tried alternate enzymes to KAPA for LD PCR in low PCR volumes say 10 ul.

Thanks

[SunPenguin]

You would have to be really, really precise to be able to sort into 2uL consistently (perhaps someone else here can chime in, but that's been my experience with any Aria sorters). More than likely, what is happening is that your cell is hitting the wall of the wells, and when you're sorting for 10 or 100 cells, the droplets are big enough to keep the cells from drying out right away.

There might be a couple things you could try. You could just raise the lysis buffer volume to 5 uL. In my experience, you usually end up yielding 2-3 uL anyway, considering evaporation and all. Even if you end up with more, you can stretch the final cDNA reaction to 12 uL without too much of a difference, in my experience.

You could also add some of the RT buffer to your lysis buffer to bring up the volume. Some of our collaborators do that (I think they actually skip the lysis buffer all together, and just sort into 5uL of RT buffer, DTT and all, without the enzyme), and it seems to work pretty consistently.

[skwek1]

Originally Posted by SunPenguin

You would have to be really, really precise to be able to sort into 2uL consistently (perhaps someone else here can chime in, but that's been my experience with any Aria sorters). More than likely, what is happening is that your cell is hitting the wall of the wells, and when you're sorting for 10 or 100 cells, the droplets are big enough to keep the cells from drying out right away.

There might be a couple things you could try. You could just raise the lysis buffer volume to 5 uL. In my experience, you usually end up yielding 2-3 uL anyway, considering evaporation and all. Even if you end up with more, you can stretch the final cDNA reaction to 12 uL without too much of a difference, in my experience.

You could also add some of the RT buffer to your lysis buffer to bring up the volume. Some of our collaborators do that (I think they actually skip the lysis buffer all together, and just sort into 5uL of RT buffer, DTT and all, without the enzyme), and it seems to work pretty consistently.

I like the last idea.
Thanks!
Serena

[Simone78]

[QUOTE=naveenrajkumar;192017]

Originally Posted by Simone78

Hi Simone,
what is the lowest total volume of preamp reaction have you successfully performed using KAPA? or when do you notice sub-optimal performance of KAPA. what was the ratio of RT in KAPA mix. I notice in your paper you use 20% cDNA in PCR volume (10ul in 50ul). Have you tried alternate enzymes to KAPA for LD PCR in low PCR volumes say 10 ul.

Thanks

we are now doing the RT in 5 ul and the preampl in 12.5 ul, half of what is in the published protocol and it works very well. The problem with reducing the preampl volume even further is that the KAPA mix comes as a 2x master mix. I also did the preampl in 10 ul (maybe 10.1 with the ISPCR primers) and it worked as well. I also tried preampl in 15, 20, 25 ul, etc. I noticed that the more you dilute the RT mix the higher the cDNA yield and the better the cDNA profile. I think this is due to the fact that the Pol in the KAPA mix is probably inhibited by the salt (or additives) present in the RT mix. I then decided to cut all the original volumes by half and that´s why I ended up with 12.5 ul.
/Simone

[Simone78]

Originally Posted by SunPenguin

You would have to be really, really precise to be able to sort into 2uL consistently (perhaps someone else here can chime in, but that's been my experience with any Aria sorters). More than likely, what is happening is that your cell is hitting the wall of the wells, and when you're sorting for 10 or 100 cells, the droplets are big enough to keep the cells from drying out right away.

There might be a couple things you could try. You could just raise the lysis buffer volume to 5 uL. In my experience, you usually end up yielding 2-3 uL anyway, considering evaporation and all. Even if you end up with more, you can stretch the final cDNA reaction to 12 uL without too much of a difference, in my experience.

You could also add some of the RT buffer to your lysis buffer to bring up the volume. Some of our collaborators do that (I think they actually skip the lysis buffer all together, and just sort into 5uL of RT buffer, DTT and all, without the enzyme), and it seems to work pretty consistently.

We have several FACS Facilities here at the Karolinska Institute that are routinely sorting in 2.3 ul lysis buffer. I know that 2.3 ul is a bit strange volume but it´s just because we later add 2.7 ul of RT mix and do the RT mix in 5 ul. We also have some international users (UK, Germany, US) who are doing the same without problems. We generally get >95% wells with a cell, although this varies with the cell type.
I think that, besides taking the time to calibrate the instrument, the model also plays a role. We noticed that the Influx (BD) is better than FACSAria III. Somehow with the FACSAria the droplet is perfectly centred for the first wells (left side of the plate) but progressively shifts so that the wells on the right side sometimes don´t receive a cell (it hits the wall, most likely). However, with patience, it is possible to get good results with the FACSAria as well. That´s my experience, but I don´t actually know if all our users who use the FACSAria had the same problems.

Interesting that your collaborators sort directly in the RT buffer! We tried but it never worked as good as with Triton. Maybe different cells require different lysis conditions.
/Simone

[skwek1]

[QUOTE=Simone78;192103]

Originally Posted by naveenrajkumar

we are now doing the RT in 5 ul and the preampl in 12.5 ul, half of what is in the published protocol and it works very well. The problem with reducing the preampl volume even further is that the KAPA mix comes as a 2x master mix. I also did the preampl in 10 ul (maybe 10.1 with the ISPCR primers) and it worked as well. I also tried preampl in 15, 20, 25 ul, etc. I noticed that the more you dilute the RT mix the higher the cDNA yield and the better the cDNA profile. I think this is due to the fact that the Pol in the KAPA mix is probably inhibited by the salt (or additives) present in the RT mix. I then decided to cut all the original volumes by half and that´s why I ended up with 12.5 ul.
/Simone

Hi Simone,
I'm a bit confused by this. You mentioned that if you dilute the RT mix more you get better amplification. However if you cut all the original volumes by half, you still have the same ratio as in your paper. Wouldn't 5ul of RT mix and maybe 25ul (final vol) of preampl (with KAPA) be better?
Thanks,
Serena

[Simone78]

[QUOTE=skwek1;192106]

Originally Posted by Simone78

Hi Simone,
I'm a bit confused by this. You mentioned that if you dilute the RT mix more you get better amplification. However if you cut all the original volumes by half, you still have the same ratio as in your paper. Wouldn't 5ul of RT mix and maybe 25ul (final vol) of preampl (with KAPA) be better?
Thanks,
Serena

Hi Serena,
you are correct. I wanted just to say that we updated the protocol and we cut all the volumes by half. Of course, the ratio RT:PCR remain the same.
Ideally we would like to dilute the RT a lot, practically we don´t do it because when processing thousands of cells (as we do) the cost for the KAPA mix and magnetic beads becomes much much higher!
Best,
Simone

[KroSeq]

Originally Posted by jwfoley

Well, SuperScript II is wild-type MMLV RTase plus point mutations to kill the RNase H domain. In principle different companies could actually be selling different enzymes with different point mutations...



Even the fancier enzymes like SuperScript III and Maxima are, as far as I can tell, just engineered to be heat-resistant. That's what makes them more processive: you could run SuperScript II at 50 °C and it would probably be just as fast, except the enzyme would degrade too quickly to be very useful. Unfortunately this engineering tends to kill the C-tailing activity for some reason. But if you aren't using template-switching, mere heat resistance is still a legitimate improvement. (For sequencing I might still worry about increased error rates at the higher temperatures.)

Hej jwfoley,
can you name references on the processivity vs C-tailing activity and the increased error rate at higher temp?

Thx!

[seq198]

This thread is indeed very interesting, and I read through all comments. As I have only limited experience with nextera, I wanted to ask whether the system is also compatible with other lib prep system like NEBnext, Rubicon or Kapa?

[hamaor]

take a look here it might be usuful
http://www.cell.com/cell-reports/abs...247(12)00228-8

[nucacidhunter]

Originally Posted by seq198

This thread is indeed very interesting, and I read through all comments. As I have only limited experience with nextera, I wanted to ask whether the system is also compatible with other lib prep system like NEBnext, Rubicon or Kapa?

Library can be prepared by shearing amplified cDNA followed by low input DNA library prep. Because of low yield most suitable option will be ThruPLEX. You may adopt the shearing method from following user manual which uses rebranded ThruPLEX DNA-Seq kit.

http://www.clontech.com/AU/Products/...10021:22372:US