Dr. Patricia de Winter: Thank you.

I'm going to be talking, following a bit on from Riel's [sp] talk that--he's just been talking about Brilliant III, to talk about some work that we've been doing with Brilliant III. And then, I'm going to finish with some work we've been doing on trying to develop some NGS standards for quantifying NGS libraries using Brilliant III.

So, I work at UCL and my David--my colleague David Sugden works at King's. And just under two years ago we formed a partnership called qStandard, and our initial aim was to run qPCR workshops. We had been doing this for a number of years anyway, but we thought why not start a partnership and do it independently? We then started to get requests for doing qPCR service work and assay design, so that's how we got into this. It took off, so we decided to continue.

Now, what do we actually do? Well, we--the bulk of our customers are actually working on gene expression. That is the most common qPCR application that we've come across. We've got the odd one that's doing weird things like exon skipping, for example. But, the bulk is gene expression for a variety of reasons, whether it's validation of micro RNA data, siRNA silencing, signaling pathway analysis, etc.

We want to quantify usually a quite large number of genes and multiple samples. So, over the past year, for example, we've actually designed 170 new assays for various people, and we've validated these. So, we do have quite a large number of assays and samples to run. So, what we need is a rapid procedure to set up sensitive and specific qPCR assays. And we also want, obviously, a dependable reagent for qPCR. And we prefer to use SYBR green just because of cost and convenience.

Now, David was constantly complaining that the robot was really quick, but the problem was that the mix we were using was taking so long that the robot would finish pipetting a run and then he'd have to wait for ages until the Rotor-Gene finished amplifying with the old mix. So, this was a constant complaint. So, it didn't take an awful lot of persuasion when Raza [sp] came along and offered him to try out Brilliant III. Now, I have to say I was less than enthusiastic because I hate changing mixes. I really do. Our old mix worked. It plodded along quite nicely, lots of assay--we'd done hundreds of assays with it. It always worked. So, I really was reluctant to change because of the work involved, trying to test it and make sure that it really would work over a large number of different assays. So--but, David was persuaded because, again, you know, he wanted the Rotor-Gene to keep up with the robot. So, the way we had been doing it--oops, sorry, was to--in order to optimize new assays that would run quickly, we didn't really want to spend a lot of time doing it, so we wanted something that always worked and we wanted a high success rate. So, generally we would use short amplicons, typically less than 200 base pairs. We tended to use conservative cycling times, so our denaturation 10 seconds and annealing 20 seconds with an extension of 10 seconds. So, that's fairly conservative. .

And we did have a reliable mix that worked very well at this--under these conditions. And it was also very stable on long-term storage. So, Brilliant III had to compete with this. So, this is what we've been--some of our--just a few of our assays in the past. You know, they've generally had good efficiency.

This one is with the old mix. You'll notice that one's actually got a low efficiency. That's a--that is a high percent GC amplicon, and with the old mix it was a little bit on the low side. We call good efficiency more than 95 percent, 95 percent or more. So, we were quite happy with it, but what did we want to improve? Well, the robot basically was fast. The assay--the run wasn't, so this is what we wanted to improve. So, the Rotor-Gene run took sixty--about 64 minutes without the melt. So, about--oh, with the melt it was about one hour and 10 minutes, one hour 15 minutes. So, faster cycling would increase our throughput. This is something that Raza touched on. So, how can we achieve faster cycling? Well, we tested this Brilliant III. We compared it with our usual mix, which we're going to call A here, and we tested it on both standards, purified PCR product, cDNA, gDNA, and RNA isolated from a variety of tissues and cells from various species.

So, you've seen this slide already. Raza showed it. But, we're using a three-minute activation. And the Taq was reported to be up to 2.2 fold higher than the wild type Taqs. It seemed to be faster. And we had all the data that Raza has just shown you. He sent that to David saying, you know, "It's resistant to inhibitors, etc." But, we want to know does it really work, because all companies claim that their mix is best. So, we went about testing it. Now, he started off under conservative conditions. This is our old mix with the old conditions in a three step assay, and tried an assay that worked well--an assay--the same assay with both mixes, with Mix A and Brilliant III. And as you can see, they're very comparable, good linearity and good efficiency.

So, he then started--he was a bit skeptical, so he just started slightly reducing the times. So, this is the first run. Then he did another one just reducing the annealing/extension, and then he kept going because he didn't want to--he just didn't believe it was going to do it for one second. And every time he did a run and got the results, I would get an e-mail or an excited text saying, "I'm down to five seconds." So, he really was getting a--he was really rather pleased with this, because it was finally down to one second and it really did work. It was good efficiency and a good R squared value with a very short run time. He managed to get the run time down to 39 minutes on the Rotor-Gene.

So, basically he discovered that it didn't--jeez, this very short run time didn't seem to compromise efficiency or linearity. So, he tried a number of other assays. Again, ran standard curves, very good R squared, and very good efficiency. This is all at an annealing/extension of one second, so we started to believe Raza's data.

The assays I've just shown you in the previous slide varied in amplicon size from 60 base pairs up to 468. I have to state this is not one that we designed. It was some primers from a student of mine. And it was originally designed for normal PCR, not for qPCR. We don't generally go up to very long amplicons, but we thought we'd give it a go anyway. That's the gene here. And David affectionately calls it Effie [sp]. He likes Effie.

So, this is some more Effie data. Here's single product, about the right molecular weight, single peak on the melt curve. And he thought he'd try--rather than just on standards, he would try it on some RT112 cells under the one second conditions.

So, I have to have a slight joke at Dave's expense here, because this is his pipetting here. This is his standard dilution. I'll show you mine in a minute. But, you can see that the efficiency is pretty similar between standard and the cDNA dilution series.

Here is my pipetting. You can see here it's better and I've eight, not one. So, these are the summary data for here. This is the standard curve with an efficiency of 98 percent, good linearity. And you can see the efficiency of all these different cDNAs that have been diluted three times, three dilutions of the original stock, have all got good efficiency. They're all above 95, so within the variability of the standard curve. So, it does seem to be performing quite well. .

So, if you prefer to measure your efficiency a different way, by looking at the individual curve, individual samples rather than using a standard dilution series or a standard curve, here is the standard, okay? So, the efficiency in the standards vary from 87.8 to 95 percent, and the efficiency in the cDNAs, which are these here, vary from 96.5 to--87.9 to 96.5. So, the variability in the cDNA efficiency is similar to that of the standard and of--so, they're amplifying--Brilliant III is amplifying the cDNA just as well as it is standards with good efficiency.

We looked at efficiency versus percent GC contents, because we've got larger number of assays. We thought we'd try--well, this is about 120, I think. I've now gone up to about 200, but I haven't put data on here. So, these are all amplified with Brilliant III, and we looked at the percent efficiency from standard curves versus amplicon percent GC. So, the GC content here, the highest--we've got two assays that are 68 percent GC, which is on the high-ish side. This gives the melt peak of about 93. And as you can see, with all of these Brilliant III amplified efficiency--efficiently, so more--equal to the more than 95 percent efficiency on the standard curve. We looked at amplicon length as well. So, that was the highest one, 486. And at one second this is. It was fine. David got a bit ambitious at this point, and he decided he was going to try amplicons more than 500 base pairs. Now, these are off some old assays that he had for PCR years ago, so he just thought he'd give them a try. And as you can see, when you're going over 500 base pairs the efficiency is starting to drop. It's pushing it. At 900, it's pretty low there, actually 70 something. So--but, this is only a one second annealing/extension step. So, what he did, undeterred, he decided to try it with a five second annealing/extension set. So, those two assays there you saw, one of 900 and one of--it was around 613. He then extended the annealing/extension, which is still pretty fast, five seconds, and then he did manage to get good results. So, he had pretty good efficiency and linearity. So, if you have very long amplicons, you can actually use Brilliant III at five seconds rather than one second.

With--how repeatable is Brilliant III amplification? This is using cDNA as a template. We've got three cell lines here, RT112, T24, and HFL1 cells. And this is his favorite. I think he likes Effie now. So, this is the longest amplicon that we had. And as you can see, the coefficient of variation for all three cell lines is very good, less than 10 percent. So, it's repeatable as well.

With gDNA as a template, again these are--we don't have many assays that amplify gDNA, so we picked two that we could find, so these two genes. And again, the coefficient of variation with gDNA is very good. It's less than 5 percent.

And then, I thought I'd try some stuff on inhibitors because of what Raza had been saying about the sodium chloride in the blood, but I thought I'd try a few more. So, this is an assay that we know is susceptible to inhibition. It's got a fairly high-ish GC content at 68 percent and 212 base pairs.

So, I added some phenol to four cDNA samples. And at this concentration, 2 percent, by the way, you can smell the phenol in the reaction. So, this is an awful lot of phenol. But, you can see that up to 1 percent phenol, there was no inhibition. But, when you get to 2 percent, then Brilliant III did start to decline in efficiency and it was inhibited.

With hemoglobin, again similar pattern. Up to--this is logged nanograms hemoglobin per reaction. So, when you're getting down to a microgram, there is some inhibition. But, again, a microgram is an awful lot of hemoglobin, and you can see that level of hemoglobin contamination as a brown discoloration of the RNA sample. So, it is quite a lot of hemoglobin. Collagen is a potential contaminant from tissue, because obviously that comes in the extracellular matrix if you do qPCR with tissue samples. Collagen had absolutely no effect whatsoever, even up to a microgram.

And urea. And I'm interested in that because I work with urine as I do uro-oncology at UCL, so I do a lot of qPCR in urine. So, urea is reported to be a potential contaminant from urine samples. This is the upper limit of normal and concentrated urine in a normal person. And that--again, there's no inhibition with Brilliant III.

So, that's what we've been doing with Brilliant III to test it. We were then happy that it really was working, so we decided we would use Brilliant III in the future. So, we did switch qPCR mix.

Now, we were approached by Sharmin Begum at CRUK, because she came on one of our workshops. And Sharmin was interested in quantifying NGS libraries. And she was completely struck by something I said on the workshop where I pointed out that, if you have a longer amplicon and you use SYBR green, then obviously your CQ will be earlier. Well, for the same amount of template, the CQ will come up earlier simply because there are more binding sites because the amplicon is longer.

So, this got her thinking. She had been trying various methods for quantifying NGS libraries. And she approached us to ask us could--would we be able to develop a method for her to use. So, this is the usual situation. You have a single target sequence within a complex sample. So, this is an mRNA sample or a gDNA sample. What we wanted to do, however, was to amplify many target sequences with an adapter at each end of the sequence. So, the orange here is the adapter. So, these are different templates, so it's quite a different situation from what we'd been used to doing.

So, the current approaches for quantifying NGS libraries are some sort of fluorescence for the detection, or using the Bioanalyzer. The problem with these approaches are that they will detect all sequences, even if you're--so, if you've got a poor efficiency of adapter ligation, then with these methods you'll still quantify those sequences that haven't had the adapters ligated.

The advantage of a qPCR approach is that you will have only sequences with both adapters amplified if you use the adapter sequence as your priming site. As far as we could tell, the only way to do that on that point was to use a standard of basically average base composition and average length. But, it was just one amplicon. It wasn't a complex mixture of amplicons, which is what we wanted to do. So, could we make more appropriate standards? So, that was our approach. We wanted to make a complex standard which would reflect the variety of amplicon sequences in an NGS library. And Sharmin wanted us to make three assays, one for 250, 350, and 450 base pairs in length.

So, she provided us with two NGS libraries and we amplified these with Brilliant III under one second annealing/extension. And we used some Illumina PE primers that were provided by Sharmin. Together, we also amplified three of our mRNA assays that were similar amplicon lengths to these three just as a comparison. So, I didn't know how concentrated these cDNA libraries that Sharmin gave me were, so I--to be one the conservative side, I diluted them fivefold before I ran them. Actually, they were really concentrated, because they came up really, really early. I just assumed she'd given me something that was almost ready to run. So, in fact, I diluted it fiftyfold and it was still coming up early. So, this lot could be slightly limited.

Our three messenger RNA assays amplified well, and we had a problem here. The NTCs all amplified, but notice that there is this increasing baseline here and amplification is quite early. It's between 20 and 25 cycles. When we ran the melt curve, we had the melt peak for the NGS libraries here and then we had a right shifted peak for the NTCs. And we thought, as one does, she must have given us contaminated primers. Don't know what they're contaminated with, but they must be contaminated.

So, what--we ran the products in a gel and we compared the--compared them with assays of known amplicon size, which was our three messenger RNA assays, and also a 50 base pair ladder. And we basically purified the 250, 350, and 450 fractions, pooled them, quantified them, purified them--purified them and quantified them, I should say.

So, we did run into a problem. The problem was that when we diluted the standards we'd made, this happened with all three assays. It was linear down to about here. So, we're starting at 10 picomolar per reaction, and five dilutions was just about okay. Efficiency wasn't great. We wouldn't consider that very good. The other two were around 90, 80-something, so not brilliant.

And we were convinced--well, I was convinced it was--she--you know, we had contaminated primers. We had a problem with some contamination and--because the NTCs were amplifying. And we'd get--that contamination was also in the low standards when we diluted them. Dave didn't think so simply because I don't usually get contamination, very rarely, and he--you know, this was massive--well, we thought it was massive contamination. So, he thought there was something else wrong.

But, nevertheless, we purchased new--brand new PE primers and we reran it. We got exactly the same result. We still had NTCs amplifying and giving that horrible increase in fluorescence in the baseline.

So, what was the solution? Well, the solution was I designed a shorter version of the PE primers. My PE primers are about--they're 60 to 66 base pairs long, each primer. So, I designed a version that was about 30 base pairs--bases from the five prime end of each primer. So, about half length or a bit less than half the length of the long primers. So, this is just NTCs now. For this one, I used two different concentrations of primers. I used the 500 nanomolar and 250 nanomolar per reaction. And I've just kept the axes on exactly the same scale for all the--for the two comparisons so you can compare them fairly.

So, there was a concentration dependency. If you use less primer, there was a less--lower increase in the fluorescent pieces. But, we still saw something in the NTCs. With the short primers, however, the increase was nowhere near as much. So, this is the same scale as this and, as you can see, that one is going up to five by about here and this one is staying flatter.

This is just an--one of our mRNA assays for comparison. You can see the baseline is close to that one. And even though there is some--looks like there is some increase in fluorescence here, there actually wasn't any product detected on the melt and neither on the gel. So, the short primers did seem to be the solution, and we ran with--ran these with the three assays. So, this is now using the short primers, from 10 picomolar down to 10 attomolar, no amplification in the NTC, very linear with a much better efficiency.

When we ran the melt, we found this so interesting effect here. And we scratched our heads for quite a while, and then my answer was that what I thought was happening was that, because you're diluting the standard--this is a complex standard. It's not like a normal qPCR standard. You are diluting out the sequences that are rarer, obviously. And when you're down at that end, you lose--you're losing the rare sequences, in effect, because they're diluted out. So, you're left with the abundant sequences, which may well have a different melting temperature than the rarer sequences, so you get this shift. Now, it's more obvious at 10 attomolar than 100 attomolar. If you've got libraries that dilute, they're not going to be much use for NGS anyway.

It's just that we--Sharmin only wanted it--us to go down five log. But, we like to go down to seven log just to show it is linear, so that's why we have gone down to 10 attomolar. But, you would be in trouble if your libraries were down at 10 attomolar. You wouldn't be able to run them. So, you know, this is actually not a problem in practice. But, why--you might ask why, if we're got a complex template when it's in high concentration, do we see only one melt peak on the melt curve. So, this is something I tried to demonstrate, why you do. I took 10 separate amplicons of various GC content and therefore various melt temperatures. So, we're going from about 79-ish to--what's the highest? About 86, 87, I think it was, something like that.

So, 10 different amplicons. I then started mixing them. So, the first one I mixed was the two that had the most extreme values. And as you can see, you get two peaks. I mixed three, four, five, etc. Five, you see they're starting to merge, five different amplicons. Mix of eight, that one started to disappear.

Mix of all of them, you're now getting--so, this is a mixture of all of these amplicons in equal proportion. You're then starting to get an odd peak that is actually at this end rather than at this end, even though most of the amplicons have got a lower melting temperature. So, that does explain why we get a nice single melt peak in an NGS library sample or in the standards, even though we know that they are a mixture of templates. So, in conclusion then, Brilliant III allows very fast cycling with reducing cycling times, which is excellent for us. We've found no loss of sensitivity, linearity, or specificity. It does seem to have quite good resistance to common PCR inhibitors. We're not claiming that obviously it's going--it's necessarily going to be resistant to all inhibitors, but it's a good start that it is performing very well with ones that are known to definitely inhibit PCR.

Enables faster cycling, therefore it's faster setup of qPCR assays and increased throughput for us. We've found that it does work well with NGS library amplicons. And the other thing we are concerned is storage. We want to be able to use it. We buy it in--we buy the 10 pack, so we want to keep it for quite a while. And we've stored it for nine months, a bit longer than nine months now, and it's still working absolutely fine. So, it does have a good shelf life as well. So, this is just a bit of what we do. We--so, we make custom standards, reference gene standards. We do contract qPCR, and we run qPCR workshops.

Thank you very much for your attention. .