Shariing our PALM experience, RNA yields from our LCM, and a Question
jm googlecalendar
Some history:
After our initial efforts were unrewarding in terms of RNA yield, we recognized that 10 microlitres of SMART-3Seq lysate solution did not completely cover the (highly convex) Zeiss LCM cap (0.2 mL tube--Zeiss # 415190-9181-000). Hence, we compared 10 and 50 ul of lysis buffer on Zeiss LCM caps at 60 degrees for 60 minutes yielded similar results: low RNA yield and cell ghosts were still visible on the caps after the digestion. Hence, we repeated these experiments using agitation (swirling inverted caps/tubes on a vortex shaker for 30 seconds every 15 minutes) during the 1 hour digestion, which resulting in higher RNA yields and left few if any cells visible on the caps after digestion.
As the 1S primer in the SMART-3Seq lysate solution dominates the Tapestation result (the expected large peak at 64 nt), we did an experiment without addition of the 1S primer. As the SMART-3Seq lysis buffer also appeared to cause considerable noise on Tapestation, compared to identical samples processed using the Qiagen FFPE RNA extraction protocol (which includes a final RNAeasy column cleanup), we ran the samples processed with SMART-3Seq over the RNAeasy column prior to Tapestation.
An experiment with six replicates was performed: 50 large circular “cookies” (1800 sq microns * 20-microns thick FFPE) were cut and catapulted by LCM from rat dorsal root ganglion tissue sections using the Zeiss PALM system and lifted onto Zeiss caps (0.2 mL tubes). After SMART-3Seq lysis (without 1S primer) with 4X agitation at 60˚C for 1 hour, and RNAeasy column clean-up (total of 14 ul volume), 4 ul of the the resulting RNA eluate was run on Tapestation.
17-Dec-21
Each sample from 25 circles (= 1800 um^2 * 20 um-thick) of
rat dorsal root ganglion cells (unstained, FFPE)
Digested by SMART-3Seq method (no added 1S primer) purified for
Tapestation using Qiagen RNAeasy column (RNA in total of 14 ul water); 4 ul of eluate used for Tapestation analysis of RNA from individual caps:
Sample Info RINe 28S/18S Conc. [pg/μl] DV200 value(%)
Smart 3Seq LCM 1 2.3 1820 72.71
Smart 3Seq LCM 2 2.2 1230 67.41
Smart 3Seq LCM 3 2.4 985 68.92
Smart 3Seq LCM 4 2.3 1090 68.12
Smart 3Seq LCM 5 2.3 843 65.16
Smart 3Seq LCM 6 2.3 563 64.16
**LCM 6 had sample loss during processing**
While the RINe values are low, they are consistent and not unexpected for FFPE tissues that have undergone fragmentation. Nevertheless, the DV200 values are rather reasonable, suggesting that the RNA is suitable for making cDNA libraries. Also note that the calculated concentration of RNA is based on 4/14 ul of eluate, suggesting that these 50 large circles contain as much as 4 ng of RNA. This particular experiment aimed to assess consistency of the method in achieving RNA yield, which seems quite good for these (relatively large) specimens.
For DV200 value discussion by Illumina, see:
The remaining question is whether RNA from similarly sampled tissues and processed completely through the SMART-3Seq protocol, i.e., with added 1S primer and without RNAeasy purification can make a library suitable for sequencing.
We are planning an experiment to sample cells from the similarly processed FFPE sections to see if the whole system is working and that we can get usable libraries. Obviously, this will depend in part on quantifying library yield (from pre-pooled samples) and estimating the correct number of PCR cycles. We also have thought to do a “dose-response” analysis of cell number to answer what might be the minimum number of cells for which the method is dependable (acknowledging that sometimes spuriously low read counts are not uncommon for SMART-3Seq).
Two questions:
(1) If, as planned, we hope to multiplex several samples, we have obtained and will use the i5 index set from IDT, however, I read your post in this Discussion Group that Illumina thinks these are unlikely to be successful if we choose to use MiSeq sequencing (for economy—based on your advice from your post 30 July 2021). Aside from using NextSeq or NovaSeq, do you have any new information from Illumina or elsewhere about trying to run the i5 indices through the MiSeq or other less expensive approach? (I am unsure I fully understand why this would be, though you say it is the unusual sequencing conditions for an experiment that has only i5 indices and single-end reads)
(2) If not, my alternative was to simply to use a library from mixed samples of SMART-3Seq processed LCM materials. I just want to know before spending considerable time, effort, and money that we actually have something that can be read and interpreted. I would value and advice you might have here.
Joe Foley
If you do have to do a cleanup anyway, you should try using a normal FFPE tissue lysis solution (maybe the one in the Qiagen kit, or see what Zeiss offers) instead of the one in our LCM protocol, which is weirdly contrived to be compatible with the downstream steps. Beyond spin columns you could also consider a bead mix like FormaPure XL, which may have a little better sample retention than columns and works well for relatively small volumes, plus you can do it in 96-well plates with with multichannel pipetting. You may also be able to recover shorter RNA fragments, which is important for FFPE, if you add isopropanol after the bead mix (with regular AMPure/RNAClean beads it's ~1.8X bead mix followed by ~5.4X isopropanol, both relative to the original lysate volume, but I don't know how it would work with FormaPure kit). Either way, if you're adventurous you could actually elute the RNA directly into the first reaction mix. Even with beads it's tricky to elute in volumes less than 10 µL, so you may need to scale up all the reaction volumes proportionally, but the beads should be inert in all the reactions so you can probably just leave them in instead of separating the supernatant on a magnet.
Aside from using NextSeq or NovaSeq, do you have any new information from Illumina or elsewhere about trying to run the i5 indices through the MiSeq or other less expensive approach?
If not, my alternative was to simply to use a library from mixed samples of SMART-3Seq processed LCM materials.
JWF
--
You received this message because you are subscribed to the Google Groups "Smart-3SEQ" group.
To unsubscribe from this group and stop receiving emails from it, send an email to smart-3seq+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/smart-3seq/898d06ec-e958-486e-a728-6dfb07284788n%40googlegroups.com.
jm googlecalendar
Thanks for sharing all this information. I would say these
samples look great for FFPE, though I'm used to clinical samples from the operating
room, where fixing the tissue carefully for research purposes is the lowest
priority.
Thanks for the feedback. I forgot to mention that for the samples in the attachment, we also included a DNAse treatment before the column purification step to assure us that what we had was really RNA.
As for FFPE, these particular specimens of nervous tissue were indeed from formalin-perfused experimental animals followed by immersion for a couple of days, so they were fixed about as well as I know how. That said, most of my career was spent with connective tissues, which live on for remarkably long intervals after death, yet their density makes them difficult to fix. Many years ago, before molecular biology was invented, I regularly used microwave fixation for both experimental and clinical samples from the OR, and that substantially improves the speed and completion of fixation for both morphology and IHC. (Heat drives formaldehyde into its gas state, which is the form which actually does the cross-linking—and powerfully inhibits RNAse).
I'm not familiar with the Zeiss system; would you be able to
email me off-list some photos of the caps you're talking about? I wonder if
there's any other way to solve this.
I’ll do that next chance I have, though we consider the problem solved in terms of RNA yield, a large benefit for the minimal cost of a few shakes during the lysis step. For those on this list, the key observation is re-examining the caps under the microscope to see if there are any cells left behind, i.e., cell ghosts = missing nucleic acid.
If you do have to do a cleanup anyway, you should try using a normal FFPE
tissue lysis solution (maybe the one in the Qiagen kit, or see what Zeiss
offers) instead of the one in our LCM protocol, which is weirdly contrived to
be compatible with the downstream steps. Beyond spin columns you could also
consider a bead mix like FormaPure XL, which may have a little better sample
retention than columns and works well for relatively small volumes, plus you
can do it in 96-well plates with with multichannel pipetting. You may also be
able to recover shorter RNA fragments, which is important for FFPE, if you add
isopropanol after the bead mix (with regular AMPure/RNAClean beads it's ~1.8X
bead mix followed by ~5.4X isopropanol, both relative to the original lysate
volume, but I don't know how it would work with FormaPure kit). Either way, if
you're adventurous you could actually elute the RNA directly into the first
reaction mix. Even with beads it's tricky to elute in volumes less than 10 µL,
so you may need to scale up all the reaction volumes proportionally, but the
beads should be inert in all the reactions so you can probably just leave them
in instead of separating the supernatant on a magnet.
Many thanks for all this, though I should be clear that the only reason we did the spun column was to verify that we were actually getting anything from the SMART-3Seq lysis. We had previous experience with the Qiagen FFPE extraction protocol, having done RT-PCR arrays, so had a kit on the shelf that we used to directly compare the same tissue (cut in half) and processed by both Qiagen and SMART-3Seq procedures. To ensure success, we used quite a bit of tissue, far more than we expected to pluck by LCM. Our initial Tapestation data clearly showed RNA in the Qiagen prep, but we did not know how to interpret the SMART-3Seq data because of the 1S primer and buffer as I described above.
In the experiments described in the attachment to this post we
still used quite a bit of tissue (25 big cookies), which is still more than we expect
to be sampling for some cells in our actual experiment. The goal of these experiments was to verify
that we were consistently getting RNA of sufficient quality as starting
material. Now that we have confidence
of this, we will do everything we can to avoid having to purify our RNA destined
for SMART-3Seq, which this forum has noted is one of true advantages of the
method for small amounts of starting material.
Note the experiment described here antedates my next experiment designed
as a dose-response curve to estimate [RNA] per cell, or at least define a lower
limit for detectable mRNA, and hopefully amplifiable, cDNA.
Aside from using NextSeq or NovaSeq, do you have any new information from Illumina or elsewhere about trying to run the i5 indices through the MiSeq or other less expsensive approach?
I think it might work on the MiSeq but I'm just not sure what settings to use. You probably have to spend a few extra cycles sequencing into the positions where the i7 index would be, which is actually just the common section of the P7 adapter and should still give detectable base calls, which you'd then ignore anyway with the modified sample sheet (that's what we do on the NextSeq but it's only 2 wasted cycles there).
Thanks, it seems doable in principle; I have just minimal experience in this area so I have begun a conversation with our NGS folks at the University of Calgary to see if they can advise me sensibly here. I’ll let you know their confidence level for success.
The question is whether it freaks out if it's the same base calls in every sample, and in the worst case maybe you just need two i7 indexes to correct that. This is actually something we might happen to test in our lab in the next week or two, but that's for experiments that are still in the testing/troubleshooting phase so it's hard to say exactly when they'll be ready. However, the MiSeq is never going to give you a realistic amount of data for a mammalian genome (unless you do an entire sequencing run for each library) so this is only useful for test runs before you continue to a higher-throughput sequencer, and you could also just take your chances directly on the NextSeq etc. If you're really hesitant you could buy a few i7-indexed PCR primers just to use for testing and then it's easy.
My only hesitation is spending six months work on this task and ending up with nothing more than a receipt for sequencing services without any interpretable sequencing data. My confidence is buoyed by your encouragement with our RNA, yet the sequencing part is what I consider a known unknown. My nature tells me to just go for it, but given the time and expense, and the fact that SMART-3Seq is new in our hands (and would be new to our sequencing staff), I would feel considerably more settled if I knew we had a library or three that we could sequence and interpret before we collect one or two hundred samples.
Once again, many thank yous for your practical advice and guidance.
John Robert Matyas, A.B.,
M.Sc., Ph.D.
Professor, Department of Comparative Biology & Experimental Medicine
Faculty of Veterinary Medicine, University of Calgary
3280 Hospital Drive NW
Calgary, AB T2N 4Z6