In this presentation, Jonathan Schmid‑Burgk, a new group leader in functional immunogenomics at the University of Bonn, introduces LAMP‑Seq, an isothermal, barcoded LAMP‑based sequencing method designed for mass‑scale SARS‑CoV‑2 diagnostics. He outlines the current testing capacity in Germany and the need for a ten‑fold increase to better control pandemics. LAMP‑Seq replaces RNA extraction and thermocycling with a single‑step LAMP reaction that incorporates a synthetic 10‑base‑pair barcode, enabling pooling of up to 10,000 samples for sequencing on small Illumina devices. He discusses key technical challenges: (1) barcode efficiency variability linked to secondary structures, (2) template‑switching leading to barcode cross‑contamination and its mitigation by extreme dilution of LAMP products before PCR, and (3) the comparable sensitivity to RT‑qPCR (detecting up to Ct 35) with high specificity (≈99.7%). Clinical validation on >600 patient swabs showed 100 % specificity up to Ct 35 and rare false positives due to contamination. Simulations indicate that a low‑cost sequencer (~4 M reads) could reliably process 100 k samples per run. He also presents a software pipeline for secure participant data entry, barcode generation, and automated plate mapping, demonstrated in a stress test with 123 volunteers. Finally, he addresses regulatory pathways in Germany, noting that university hospitals may use self‑validated methods without formal approval, though commercial kits would require conformity assessment. He outlines next steps: larger stress tests, ring‑trial validation, faster sequencing protocols, and preparation for future pandemics.
Chapters
Thank you once again. So in, in the interest of time, we have to go on, and our next speaker is Jonathan Schmidburg. He recently came back to Bonn. I actually do not know if he has been there before, but from the Broad Institute, he is— he became their professor and group leader for functional immunogenomics at the Institute of Clinical Chemistry and Clinical Pharmacology. So, and his presentation will be about scalable SARS-CoV-2 diagnostics using LAMP sequencing. Jonathan, if you can share your screen. Yes, you are doing so. Perfect. You may start, the stage is yours.
Hello everyone, thank you Peter, thank you Alexander for the invitation. It's a great pleasure for me to be invited to present some of our unpublished data here on a mass-scale diagnostics method that we call LAMP-Seq. So Germany has currently a capacity to test roughly 1.5 million people for a SARS-CoV-2 infection per week, which is pretty good actually. But there are simulations, mathematical simulations by epidemiologists that say if you could increase that number, let's say by a factor of 10, this could actually be a very important contribution to suppress a pandemic like this one. Maybe not the current one, but maybe in future so as a large-scale testing technology could make a contribution. So when I started my lab in June at the University of Bonn, we quickly came together with a group of young PIs actually, Kerstin Ludwig, Ricarda Schmidthausen, and we sat together and asked how can we actually contribute to maybe increase that testing capacity. So the typical testing in Germany, the gold standard, is you rehydrate a swab sample, typically from the throat, you extract RNA, and then you do an RT-qPCR. This is very scalable and very sensitive and also very well established at many centers. For example, here in Bonn, I think the capacity is somewhat like 2,000 samples per day. The problem is that it's not unlimitedly scalable. On the left side here, you see an RNA extraction robot that is actually being used here in Bonn. On the right side, you see a typical thermocycler, and there is only a limited number of those. So we quickly turned to what we like most in the lab, which is sequencing devices. I'm showing here really the smallest ones that you can get, and Actually, there's a lot of these small sequencing devices in Germany. For example, the right one here, the NextSeq, I think there is roughly 200 machines all over Germany, many of them in academic institutions. So we asked ourselves, how can we actually leverage the power of these machines, each of them producing millions of sequencing reads per day, and to actually do mass-scale testing for SARS-CoV-2? So if you look at an amplicon-based sequencing outcome of one of these machines for SARS-CoV-2, Typically, it looks like this. Some of your sequencing data maybe cannot be used because it's covered by your primers. I've grayed out those sequences. But then in the middle, you have kind of a signal sequence that is a typical signature for a viral infection because the probability of getting the right, let's say, 10 base pairs just by chance is very, very low. So the specificity of a sequencing-based diagnostics method should be very high. And so the first thing we turn to is how can we actually scale up the production of sequencing libraries from, let's say, hundreds of thousands of swabs per day. And that's actually a much bigger challenge than getting enough sequencing power. So the design that we chose is shown here. We not only sequence part of the viral genome, we also introduce a barcode sequence as shown here in the middle. In this case, just a 10-base pair synthetic DNA sequence of our choosing that can be used to pool a lot of samples together and analyze them in a scalable fashion. The other thing that we wanted is we wanted to develop a method that is not relying on RNA extraction or thermocycling. So we chose a LAMP-based amplification and barcoding method, which had not been deployed before. So here is our outline. I will skip over all the development steps. So the final protocol only uses one single pipetting step. And I think that's what actually makes it scalable. You start from an oropharyngeal swab that goes into a robotic-compatible deep well plate with lysis buffer. You heat that up once, you only pipe it once into a LAMP reaction that's shown here in the middle. The LAMP reactions are pre-made to contain different barcodes for each sample. And after that has been incubated in a water bath for a bit over an hour, you can just pool together up to 10,000 samples and put them together on a sequencer. The sensitivity of this LAMP-based method— and we worked a lot on the details on that— but it's right now, it's in the similar ballpark than the RT-qPCR method. Based on the assay. It's not incorporating the enrichment of RNA that you get by extracting RNA that gives you a higher sensitivity, slightly higher sensitivity for the qPCR-based methods. But without RNA extraction, we're in the same sensitivity range as a qPCR detection method. Very interestingly, while I think a lot of you use barcoding every day in your lab procedures, for multiplexing sequencing reactions. So we thought that you can probably use any sequence you want as a barcode. So we just, as a start, we got 480 barcodes and just tested them all. And to our surprise, very different from what we typically see by PCR-based library preparation methods, we see that our isothermal LAMP amplification method is actually not agnostic to the barcode sequences. We see that there is actually a large spread in efficiency of incorporation of individual barcodes, which is totally critical to know for any diagnostics obviously, because otherwise you might just lose samples because you're using the wrong barcodes. If you look closer to the ones that actually perform worst, we found that actually they all contain a little motif GTCC, which actually matches the reverse complement of the FIB primer. LAMP has 6 primers, not only 2 as in PCR. It's very interesting because the LAMP is constant at 65 degrees, so in a way it makes sense that if you don't melt your primers in every cycle, there is a much higher chance that if your primers tend to form a secondary structure, like probably those those guys here, then you're ending up with a barcode sequence that cannot be amplified. So very importantly for this method, all the barcodes we're using for clinical diagnostics have to be tested physically before being deployed. Another problem we turned our attention to was template switching. Like in many sequencing-based methods, we were interested in how specifically can we actually barcode samples. Do we get any reads with the wrong barcode? And that could later lead to a wrong diagnostic result. So for asking that question, we inoculated some of our RT-LAMP reactions with RNA, whereas other barcodes were not inoculated. And as you can see here, the ones that had not been inoculated also gave rise to in the range of 1% of the reads of the positive reactions, which is actually not sufficient for diagnostic method. But we reasoned that one possible explanation for this could be that the barcoded primers that you're using in LAMP are still present when you pull the LAMPs and put them into the PCR reaction for prepping the library. So one option here would be that these primers, the barcoded primers, would, would still act as a PCR primer in the second stage and could give rise to wrongly barcoded samples. And if that was the case, we could just dilute our samples, because LAMP reactions typically produce an enormous amount of DNA, typically 10 times higher than a PCR reaction by concentration. So we should still have enough template DNA for our PCR. But we should dilute the primers and have less template switching. And that's actually the case. We can dilute our pooled LAMP reactions 1:1 million. We still get the same numbers of saturated PCR reads, obviously, but we eliminate the problem of barcode swapping here. Finally, we validated this protocol with clinical samples, and I have to say, very grateful here at the university hospital for the support we got from Martin Exner and Markus Nöth and others. Who gave us access very quickly to over 600 actually clinical swab samples where we could compare the gold standard RT-qPCR method to our new method. And the results were quite encouraging. We see 100% specificity up to a clinical CT value of 35. One might argue about what the actual cutoff should be for detecting infectious people, but I think 35 is pretty safe from what I understand from virologists to say if someone is infected or not. And then we have also a very high specificity. It's not 100%, %. We sometimes get a false positive, very rarely, but I'm sure we can still solve that problem as well, which is not because of sequencing. It's more because of some physical contamination problems that we're still tackling, but the specificity is still very high, 99.7%. Finally, we wanted to assess how many samples we could actually multiplex in a single sequencing run, and so we turned to actually the smallest sequencing device available to my knowledge right now, which is only $30,000 or so,. And it's very easy to deploy because it doesn't have any complicated optics inside. So it could even be deployed remotely in the field pretty quickly. And so we just downscaled, subsampled the data that we got from our previous results to simulate how many samples actually could be reliably detected. And the results are quite encouraging. Even this very small sequencer, which makes roughly 4 million reads, can still analyze 100,000 samples based on this simulation. Per run or per day. So finally, and this actually was a much bigger challenge than we would have anticipated, Kerstin Ludwig actually led the way to develop a way to deploy this method in scale. Because it's not only about turning the swabs into libraries, it's also about acquiring enough swabs and getting the private data of people into a safe database. So we developed this software, which we will make open source, obviously, which starts with a website where people can enter their private data and choose a time slot from our database. Then they create a barcode that's shown here in the middle. They print this barcode by themselves. So they come to the swabbing site with a printed barcode that contains all private data. Then this barcode is scanned and the software that you see on the right will actually talk to the medical personnel and instruct them exactly where in the robot-compatible plate the swab has to go. So it's very unlikely that confusions would arise during this swabbing procedure. And we actually had our first stress test just 2 days ago where we tested— we just asked our colleagues here at the institute to participate and successfully tested 123 people. So finally, I would like to highlight the next steps. We want to test the scalable swab acquisition and really in more stress tests and bring it to higher scale. We want to make sure the hygiene and the logistics is all solved. We also want to participate in ring trial validations, which is important for actual clinical application of this method, and we want to speed up the method. It's not a quick test, I would say we have to do a sequencing run. As most of you know, there are some sequencers that are faster than others. The iSeq is not the fastest one, but there might be solutions to even do a very short sequencing run in just 3 or 4 hours, for example, on a MiSeq, if we have a smart design of our sequencing primers and the barcoding strategy. And finally, the biggest goal of all, and I think the whole team here agrees since we're doing this since a few months, we still want to get this going in this pandemic. It might be unrealistic, and obviously I hope that the vaccination strategies will maybe obsolete need this method, but I would argue if it takes a year to vaccinate the whole population, or maybe in other countries it might take even longer, we hope that we could still contribute something to large mass-scale testing procedures and also be prepared for a future pandemic. So most of our results are published as a preprint, and I hope that also a paper will come out soon. And I'd like to thank again especially Kerstin Ludwig, Ricarda, Thomas, Ines, who makes all our enzymes ourselves so that we're independent from suppliers, Per Hoffmann, who is our robotic expert, and Katja Blumenstock. And because it's a young lab, I might take the chance to invite anyone to apply if you want to do a PhD or
postdoc. Thank you. Jonathan, we thank you. It was, it's really an encouraging development. My question would be, what do you think will be the obstacles to get it into regular diagnostics processes in
Germany? Yeah, so interestingly in Germany, the approval process for diagnostic methods is very different from the FDA in the US where you have to apply with your data and they give you kind of an authorization to use your method. In Germany, there is a lot of responsibility on the side of the developer of the method. So what's very important is if you want to deploy your method by selling it, like by selling a kit or so, then you have to go through a conformity, which is not trivial to do for an academic institution, but it's possible. But what's even more interesting is that as a university hospital, you're even allowed to use a self-made method if you make sure it's actually working, you document it without approval. That's very different from the US. So we hope that once we can have convincing data and all the steps of the method are actually shown, that we will be able to actually use it here at the university hospital within a few months of
time. Okay, thank you. There is one question here from Maria Bukhari again talking about the sensitivity. Is LAMP-Seq still useful to detect at very early
stage? Yes, so as I mentioned before, the assay sensitivity of a single reaction is comparable to qPCR. So yes, it is very sensitive. It's a nucleic acid-based detection test, and it is probably orders of magnitude more sensitive than an antigen quick test. Antigen quick test, right now the numbers circulating in the community are that you can detect people up to a CT value of between 27 and 30, maybe 33 with new developed antigen tests. But as we have shown, we have a very good sensitivity up to 35. It's very unscientific to talk about these CT values, but unfortunately that's kind of the currency that is circulated these days. So I would argue that our method is very sensitive also for asymptomatic people. It would be useful for screenings, and that is what it discerns it from, separates it from antigen quick
tests. Okay, thank you very much, Jonathan. Much success for your further developments in your group. So we have now
