RNA-seq from single cells?

[SUGU]

Hi Simone,
I have been using your Smart2-seq protocol on single mouse neurons. Based on Bioanalyzer and Fluidigm Biomark with 96 primers looks promising to move forward for sequencing.

I wonder whether you have used ERCC as a internal controls, if so what concentration works best in your hands for single cell.

I curious to know, (since Linnarsson is in your neighbourhood) have you tested your TSO with UMI for smart2-seq. If so, what is your experience....

Best,
Suguna

[jparsons]

From the discussions i've had with C1 users, Spike-ins for single cell are unfortunately very cell-type specific - you want to maintain a 3-5% amount of external RNA to cellular RNA, but the amount of cellular RNA (and lysis efficiency) varies between cell types and even between cells. For practical purposes, though, two papers both used a 40,000 dilution from the Ambion product on the C1.

If your cells aren't too rare, you can try to calculate an average RNA mass from bulk. If they are, you may have to do a test run or two to get the ratio right.

The good news is that the ERCCs are useful once you get them spiked into your cells ; you can compare the technical variability of ERCCs to that of the measured transcripts/genes to separate biological variability from random noise and calibrate to how much mRNA was present in each cell. Another (slightly off-label) use is also determining the efficiency of reverse transcription.

[SUGU]

Thanks jparson for your reply.

I have used C1 with ERCC dilution you have mentioned above.

C1 works best for cultured cells in terms of capturing almost 80-90% of the wells. It is cost effective and more cells to analyze.
Whenever I load my primary cells on C1, only 40-50% of the cell get captured. so my cost of prep is high.
In addition to that you cannot use different experimental conditions in the same chip. Unless you have two many C1 machines.

That is the one of the reasons I moved to alternate manual methods.

With Smart2seq, I could able to FACS different cell types and experimental conditions on the same day. Since I want to take advantage of ERCC, I want to know what would be the ideal ERCC concentration for single cell sorting conditions..
Now I am using 10 Million final dilution per reaction. I wonder whether this would be enough....

Based on Simone's bioanalyzer profile I could nt say whether he used ERCC's and I dont know how he determines his RT effieciency in different cells run at the same experiemental conditions or how to determine the technical variation across different samples...

Best,
Suguna

[Simone78]

Originally Posted by jparsons

From the discussions i've had with C1 users, Spike-ins for single cell are unfortunately very cell-type specific - you want to maintain a 3-5% amount of external RNA to cellular RNA, but the amount of cellular RNA (and lysis efficiency) varies between cell types and even between cells. For practical purposes, though, two papers both used a 40,000 dilution from the Ambion product on the C1.

Totally agree here. One has to optimize the amount of ERCC case by case.
For "big" cells (cells in culture: MEF, HEK293T, C2C12, P19, HeLa, etc) we use 0.1 ul of a 1:40000 dil., which gives a final conc in RT of 1:4 million. However, when working with extremely small cells (T-cells, for example) we are using a 1:40 million dil., which is still too high because we get 5-10% (!) reads from the spike-ins in the end. Maybe 1:100 million is the right conc?
Again, as parsons said, there other factors that play a role: lysis efficiency, intracellular variability, cell cycle stage, how "stressed" your cells are, lysis buffer you are using, etc etc

Originally Posted by SUGU

I curious to know, (since Linnarsson is in your neighbourhood) have you tested your TSO with UMI for smart2-seq. If so, what is your experience....

as far as I know they haven´t tried it yet but it´s quite some time that I don´t talk to them. I think the idea was to run the Smart-seq2 on the C1 but so far there had been other problems, such as the fact that the PCR protocol couldn´t be changed. And this was key to get a good preampl because KAPA HiFi requires 98 degrees for denaturation and higher annealing T, while Advantage 2 requires lower T both for denaturation and annealing.

Best,
Simone

[SEQquestions]

Hi,

For those of you using ERCC spikes in your single cell RNAseq on the C1, do you have examples of bioanalyser traces for negative controls? (i.e. capture sites with no cells in). We are trying to see what a negative control should look like as the ERCC spikes will be amplified so should show on the trace regardless.

[jwfoley]

I curious to know, (since Linnarsson is in your neighbourhood) have you tested your TSO with UMI for smart2-seq.

It wouldn't work that way because then only one fragment from the cDNA would contain the UMI. However, if you don't constrain yourself to resequencing the entire transcript, then you can get a true molecule count, and that's what Linnarsson's group did in doi:10.1038/nmeth.2772.

[dtm2451]

Hi Simone,

Thanks very much for all of your comments and recommendations in this thread.

I am working to test/optimize the Smart-seq2 protocol for a Tcell application and am wondering how you blocked your TSOs. You mentioned iso-nucleotides earlier on in this thread. Is this what you were referring to?
iso-dG: http://www.idtdna.com/site/catalog/m...s/product/2188
iso-dC: http://www.idtdna.com/site/catalog/m...s/product/2187
description: http://www.idtdna.com/Site/Catalog/M...ons/Category/7

I have little experience with such nucleotides in the past. Is the idea that iso-dG and 5Me-iso-dC pair with each other and not with regular A,C,G,T, thus the RT stops when it comes across an iso-dC or -dG because its counterpart does not exist in the dNTP mix? Is concatemerization prevented then by obstructing the RT before it can add another round of untemplated Cs after the copying the TSO?

Best,
Dan

[jwfoley]

There is an extended discussion about oligo blocking on previous pages in this thread. Other groups are using 5' biotin as a cheaper way of doing the same thing, but it sounds like you understand the principle. See doi:10.1186/1471-2164-11-413 for more.

[dtm2451]

Thank you jwfoley. I am actually thinking of trying both and testing biotin vs. iso vs. unblocked. But, I'm not sure how to get an oligo with both iso and LNA. Looking back, you actually mentioned this issue before. Were you able to find a place to order an LNA-iso combo or did you switch to biotin and avoid the issue? If not, has anyone else been able to order an oligo containing both locked and iso nucleotides? And if so, where from?

[jwfoley]

No, I never tried iso because biotin completely eliminated the hedgehog in my tests. However, if you talk to Exiqon's sales reps, they seem flexible about ordering things that are "off the menu" (since they apparently just subcontract their orders out to companies like IDT anyway).

[Simone78]

Originally Posted by dtm2451

Thank you jwfoley. I am actually thinking of trying both and testing biotin vs. iso vs. unblocked. But, I'm not sure how to get an oligo with both iso and LNA. Looking back, you actually mentioned this issue before. Were you able to find a place to order an LNA-iso combo or did you switch to biotin and avoid the issue? If not, has anyone else been able to order an oligo containing both locked and iso nucleotides? And if so, where from?

I ordered an iso+LNA-modified TSO from Eurogentec (http://www.eurogentec.com) and it worked well. However, the "performance" (number of genes detected at different RPKM levels, for example) was always slightly lower than the standard LNA-TSO and I didn´t continue further. It actually never got rid completely of the concatamers, even though there is a published paper claiming it will. Biotin could be an alternative solution.

[JJMS]

Did somebody ever use the Ampure purification in between the 1st strand synthesis and cDNA amplification in the SMARTseq2 protocol? (like Clontech has standard in it's ultra-low protocol)

[ustar]

Originally Posted by MMaria

Interesting. Does anybody knows how much does it cost or which labs already have it?

We already got C1 and run twice. cDNA harvested from C1 was around 0.2~0.5 ng/ul and we basically got 13ul for each sample. The libraries for single cell samples run well on Miseq and we get 3M reads per sample when we pooled 6 per run.

[seqgirl123]

Is anyone quantitating the amount (ng) of the total RNA in your samples prior to using these single cell kits? Is it necessary for these kits, especially the Fluidigm?

[wacguy]

Hi seqgirl,

I use a rough estimation from the bioanalyzer (the Qbit minimum RNA amount is too high for such low quantities); if you have single cells RNA, I don't think there is any way to quantify it (or even get the RNA tracks w/ the bioanalyzer).

Guy

[nucacidhunter]

If these kits are being used for low input, it is a good idea to QC input RNA because kits are expensive. When using single-cells in C1 it is physically impossible to QC RNA and the approach is to check success after harvesting amplifid cDNA. The same goes if single-cells are processed manually.

[ChristmasSunflower]

Does everyone here do cell-sorting and cDNA synthesis on the same day? has anyone here tried to use pre-frozen sorted single cell with SMARTSeq protocol?

We tried but only a very few of reactions worked so far. Wanted to know where we can troubleshooting. Really appreciate your input.

Thanks

[Dimitri Szymczak]

Originally Posted by CTC

Thanks for the article. But without detailed protocol it is hard to replicate the experiment.
I have tried SMARTer PCR kit but it did not work in my hands. The expected yield for a single cell is 7 ng, the negative control yield is similar.

For 1ng to 5ng total RNA input there is the TotalScript developped by Epicentre (quite similar to their Nextera). The protocol is already optimised and well defined .

The procedure takes 5 hours - librairies are oriented.
It works on Eurcaryotes. There are actually 3 methods depending on the goal of the project. One is based only OligodT and you would see only 5% of reads with rRNA.

[lacarb19]

Originally Posted by ChristmasSunflower

Does everyone here do cell-sorting and cDNA synthesis on the same day? has anyone here tried to use pre-frozen sorted single cell with SMARTSeq protocol?

We tried but only a very few of reactions worked so far. Wanted to know where we can troubleshooting. Really appreciate your input.

Thanks

Hello, we've tried it with Smartseq2. We sorted two plates, one that we froze down and another where we did cDNA synthesis the same day. Oddly, the sample when cDNA synthesis is done the same day gave more false positive expression (based on our known cell type) than the samples that were frozen first. We didn't follow this up but just used this experiment to justify the fact that we incorporate freezing first before doing cDNA synthesis (usually next day).

[lacarb19]

Hello all, thanks for all your thoughts here. It's helpful.

Just curious about how well your bioanalyzer traces correlate with actual sequencing data. With our samples, we can't see much of a correlation.

For example, in the attached pdf, sample D2 appears to have some concatamer and that sample mapped poorly to mouse genome (8%) and we detected very few genes (less than 200). Sample C8 seems to have a higher amount of product (between 500 and 10,000 bp) and concatemer. But this sample mapped 50% to mouse genome, and we we detected almost 5000 genes. Sample G5 has no concatamer, primer dimer (at ~100 bp), and noticeable product, but had 14% mapping and 600 genes detected.

In short, we can’t seem to find a correlation between what we observe on the bioanalyzer trace and what we’re seeing with sequencing, at least by mapping % and genes detected. We’re actually considering skipping bioanalyzer since sequencing costs on a miSeq for 96 samples (~$1000) is about equal to two Bioanalyzer chips (11 samples per chip) anyway. Anyone have any thoughts on this?

Thanks.