Learn how Biomek automation of NEBNext RNA library preparation enables high-throughput, reproducible production of libraries with high yield and complexity. In this webinar Zach Smith from Beckman Coulter describes strand-specific RNA library preparation using the NEBNext Ultra Directional RNA Library Prep Kit for Illumina and discusses Biomek automation of this method.
My name is Zach Smith, and I'll be talking about the automation for the NEBNext Ultra Directional RNA Library Prep Kit for Illumina, constructing effective strand specific RNA-Seq libraries from low input samples.
So, we have a number of RNASeq library construction workflows already automated on our various liquid handling platforms, including the TruSeq RNA version 2, the TruSeq Stranded mRNA, which is an Illumina Qualified method, and the Epicentre ScriptSeq Complete Gold Low Input kit. We're also automating the NEBNext Ultra Directional RNA kit, which is what I'm going to be talking about today.
The question is, is why would you bother doing directional or strand-specific RNA-seq in the first place? Well, in strand-specific RNA-seq, we're actually maintaining the originating strand information during the course of library construction, using a dUTP replacement in the cDNA synthesis. What that does, is it allows you to identify antisense transcripts, determine the transcribed strand of other noncoding RNAs, and determine expression levels of coding and noncoding overlapping transcripts.
Additionally, strand specific RNA-seq allows for easier mapping to reference genomes, which can lead to a substantially enhanced value of your RNA-seq experiment, and allow you get more data for less money.
So, directional or strand-specific mRNA, as far as NEB's goals in making this kit, was to produce high strand specificity, high yields, a fast, a simple protocol, and low cost.
They used a dUTP replacement strategy, which they have licensed from the Max Planck Institute. It's a published and highly regarded method, in which what we're … Their kit enhancements include a faster, streamlined workflow, and novel reagents, resulting in the NEBNext Ultra Directional Library Prep Kit for Illumina.
This is a schematic of the workflow. We're going to start with either total RNA, or poly-A enriched mRNA, or ribo-depleted RNA as starting materials. We'll then go through RNA fragmentation and random priming, and first strand synthesis as normal, using a standard dNTP mix. During the course of second strand synthesis, we're going to use a dUTP replacement to mark the second strand, and then we'll follow through that with end repair A-tailing, and adapter ligation.
Adapter ligation for the NEBNext Ultra Directional Kit involves using a proprietary hairpin adapter, that is then converted into a standard Y-shaped adapter after uracil excision with the user enzyme from NEB. We then do PCR enrichment, and then we have a completed library that's able to be sequenced on all Illumina sequencing platforms.
The NEBNext Directional RNA produces highly strand-specific libraries, you can see here some data from New England BioLabs, where they were comparing the NEBNext Ultra RNA, which is a non-directional kit, and their Ultra Directional with, and without, actinomycin D. What they were able to show is that they're strands kit produces its best work when working in conjunction with actinomycin D, which is known to block DNA-specific priming of reverse transcriptase. But even without that, it still does a fairly decent job of producing libraries that are effectively stranded, as compared to the non-directional kit that you see on the left here.
Here's just a basic map of the workflow, again, just from an automation standpoint. The blue boxes indicate places where your inputs are, and then we have various stops. On the left hand side, you'll see the cDNA synthesis workflow, where we'll start with either 50 microliters of total RNA and go through mRNA purification using the poly enrichment module from NEB, or we'll start with 5 microliters of pre-purified mRNA or ribo-depleted RNA, and proceed directly into mRNA fragmentation and priming.
We'll then go through first and second strand synthesis, and then double strand cDNA cleanup, and then that's the completion of the cDNA synthesis workflow. A user can stop here, or they can choose to go into the RNA construction workflow, which is on the right, starting with our purified double-stranded cDNA, we'll then do end repair and A-tailing in a single step, and adaptor ligation and cleanup. We have the option in the method of stopping at that point, or proceeding on to PCR enrichment setup and running through that, where we can also stop the method after PCR enrichment, or we can clean up the final enriched library and move on to library characterization and sequencing.
Just some numbers we were able to put together. For the cDNA synthesis method, for up to 96 samples, we were able to complete that entire workflow in 5 hours and 9 minutes. For the library construction workflow, 4 hours and 19 minutes, for total of 9 hours and 28 minutes for 96 individually barcoded, strand-specific RNA-seq libraries.
Those numbers are substantially less than doing 96 manually, which we estimate will take up to 21 hours, which would result in a lot of annoyance on your technicians' part. We're offering some substantial time savings for this automated method.
Our automated method on the Biomek FXP includes an HTML-driven user interface that allows any user the ability to just go ahead and run through the method, without knowing a good deal of the specifics of Biomek programming or the method.
On the left, you'll see the cDNA synthesis user interface, and we have a number of options here, including running them … So we have an option of running the method in training mode, so if you're starting with a brand new technician, one of the things you can do is you can click this box and it'll run the method in simulation. That will allow you to train your technicians more effectively.
When you're running it in live mode, you can enter the number of samples, up to 96, and we also give you the option of whether or not you're going to use on-deck or off-deck thermocycling. If you choose to do on-deck thermocycling, we have an integration with the Biometra TRobot, that will allow you to do that. If, however, you choose to go off-deck, the method will then create several pause steps, allowing you to take the plate off and put it on your thermocycler of choice.
This drop down menu will control which workflow you're going to. So, this will be effectively the first one you choose, whether you're going to do cDNA synthesis, or you're going to do library construction. If you're doing cDNA synthesis, then you have the option of checking to perform the poly-a enrichment. If you were going to start with mRNA purified or ribo-depleted messenger RNA, you would simply un-check this box.
The library construction UI, which I'm highlighting now, also has a training mode option, and a number of samples, and the option of doing on-deck or off-deck thermocycling, and several other options in the library construction workflow, including a user pause to add the adapter plate to the deck. You have the option of deploying the adapters in a custom-designed labware, in this case we're using a Bio-Rad Hard-Shell 96 well plate. Whether or not you wanted to perform PCR setup, if you're performing PCR setup how much of a volume of pre-enriched library you're going to use as a template, and whether or not you're going to perform PCR cleanup. It's a very intuitive user interface that we think will make operation of this method quite simple.
This is a picture of the cDNA synthesis deck setup, so as you can see we have a 96 channel wash station here, an orbital shaker under this plate, our integrated TRobot on this deck is right here. We have a Static Peltier to maintain our master mixes at 4 degrees, and various ALPs (automated lab positioners). And so, this deck is actually reflected in the deck setup of our Biomek FXP, which is behind me.
If you're going to be doing on-deck thermocycling, the method will add a stackable lid that will then be put on the plate that you're doing the thermocycling to. If you choose not to do on-deck thermocycling, then that piece of labware will be eliminated from the deck setup. To its right, we have an AB1127 plate stacked on top of our Bio-Rad Hard-Shell.
The library construction method takes advantage of the same deck setup. So, while we have a 96-channel wash station, orbital shaker, and Static Pelt, if you're doing on-deck cycling we'll have the TRobot, and we have a variety of stacked plates. If you choose to leave the adapter plate on-deck, and not do the adapter pause option, then the deck will be populated with an ACME cold block for keeping the adapter cool. If you choose to do the adapter pause, then that will just simply put the adapter plate in place, without the ACME block.
Just going to run through some quick experimental data that we generated in-house. We ordered Agilent Universal Human Reference RNA, and tested it out with 25 nanogram and 100 nanogram total RNA inputs. We had diluted the NEB adapter down 1 to 40, after doing several titrations, and these graphs below right here are basically just how our plate was laid out.
This is some Bioanalyzer data of our Universal Human Reference RNA libraries at 25 nanogram and 100 nanogram total inputs. As you can see, our yields were quite good. The size distributions of the libraries was quite typical for an RNASeq library, and we also do not note the presence of the adapter dimers at 120 basis. So, we're very happy with how these libraries look.
We then ship the libraries to New England BioLabs, where they did qPCR quantification of the libraries, and our yields are stacking up quite comparably to their manual libraries that they made in-house. At lower inputs, our standard deviations of our qPCR results were actually tighter than what they were able to achieve manually, so we were very pleased with those results.
And then, NEB conducted a single MISeq 2 by 100 cycle paired-end run, loading 8 picomolar onto the flow cell on the eight libraries that we generated on this automation run, and we had very high pass filter rates, north of 90% as you can see here.
This is just some fast QC data of the quality of our libraries on a per-base basis. What you're seeing here is the average QC score along every cycle of the sequencing run for both our 25 and 100 nanogram inputs. As you can see, we're above Q30 in both cases.
Just some more fast QC data, showing the mean quality scores for both our 25 nanogram and 100 nanogram inputs. We're generating very high quality data, but is it reproducible? One of the great selling points of our Biomek is the ability to create up to 96 libraries simultaneously.
What we're looking at here is the reads mapped back to the reference, using TopHat2, and then put through Cufflinks, to look at transcript expression correlation for both our 25 nanogram and 100 nanogram RNA inputs. As you can see, on the graph on the left, we're looking at the logFPKM of two of our 25 nanogram inputs, and you can see we have very high R squared correlations of up to point 99. Also, on the right, you'll see that our 100 nanogram libraries also have very high correlation within the groups.
Looking at the sequencing read composition, we're overwhelmingly looking at coding region with flow percentages of ribosomal RNA. So, we're actually targeting the parts of the genome that we're interested in getting at, which is the coding regions.
Our transcriptome coverage is very uniform across the length of the transcript, so we're not seeing any 3 prime or 5 prime bias in our samples among our various replicates.
We do see a drop in complexity for the lower input samples, which is not terribly surprising. So, you'll see on the left are our 100 nanograms input for replicate 1, 2, and 3, and on the right are our 25 nanogram inputs for replicates 1, 2, and 3. You can see that our percent duplication among those lower input libraries is indeed a little bit higher, but we feel like this is very solid data.
The other questions we wanted to ask is, as far as comparing our 25 nanogram and 100 nanogram inputs, well how do those libraries stack up to each other? This graph shows our 25 nanogram FPKM information on the Y, and then the 100 nanogram on the X, and we're seeing very high correlation between those two libraries. Regardless of whether or not it's 100 nanogram input or a 25 nanogram input, we're still seeing very decent correlation between the two samples.
So, we also wanted to look at local strand specificity. This is a GABARAPL 1, and just an IGB graph showing that in Read 1 all of our reads are pointing one way, and in Read 2 all of our reads are pointed the opposite direction. If this was a non-stranded library, we would see a roughly even mix of read directionalities. As you can see, we're not effectively seeing that, so we're confident that these are highly strand-specific libraries.
In summary, the automation of the NEBNext Ultra Directional RNA Kit provides the ability to build up to 96 high quality, directional RNA-seq libraries for sequencing on any Illumina platform. We have a highly flexible user interface that allows you to choose between the cDNA synthesis and library construction workflows, to vary your sample inputs as far as number of samples, and to get different method outputs depending on your requirements. And, our optimized pipetting allows users to apply low concentration sample inputs for directional RNA-seq, which allows you to generate high-quality technical replicates, or to use precious or limited samples for next generation sequencing.
With that, I would like to thank our collaborators in New England BioLabs, including Daniela Munafo, Deyra Rodriguez, Eileen Dimalanta, Brad Langhorst, and Stephanie Graber, as well as Mary Blair, Alisa Jackson, Julie Moore here at Beckman Coulter.