Video Chapters
0:19 - Why Metagenomics Became Inevitable for Diagnostics
1:44 - What Sequencing Enables Beyond Culture and PCR
2:36 - The Hidden Bottleneck in Clinical Metagenomics
5:03 - The Accidental Discovery That Changed Host DNA
9:55 - What It Takes to Make Metagenomics Work in Hospitals
Overview
Rapid metagenomics could revolutionize how we diagnose infections—but getting it into hospitals took a decade of breakthroughs and setbacks. Professor Justin O’Grady was one of those leading that charge, by moving from PCR to sequencing at the University of East Anglia to crack a critical bottleneck. A 2015 conference conversation sparked an experiment that succeeded thanks to an unnoticed error—the turning point for rapid clinical metagenomics.
In this video, Justin traces the journey from lab discovery to revolutionizing patient care. After proving his method at St Thomas's Hospital in London, Jonathan Edgeworth's team further refined it into a respiratory service now used across UK ICUs. Today at the Ellison Institute, Justin is pushing to make metagenomics the standard for pathogen detection.
If you’d like to hear how the method Justin introduced to the Edgeworth team was further refined into a validated approach now being implemented across UK hospitals, join our upcoming webinar.
Transcript
Part 1 — From PCR to Sequencing
Tell us about your move into metagenomics
I started off in the area of, PCR and qPCR. And it wasn't until that I took my first significant academic role at the University of East Anglia in Norwich in the UK that I started to think about metagenomics. The person who hired me is a professor of medical microbiology called John Wayne. He hired me for my diagnostics expertise, but he was very interested in sequencing, and he was keen that I would apply sequencing to diagnostics. And he was aware of the work that was ongoing at Oxford Nanopore Technologies, shortly after that famous talk by Clive Brown at AGBT where he introduced the MinION sequencer.
And the exciting thing about that was that you were going to have a sequencer that was able to produce data in real time and move it on to a laptop in real time. And that was going to change how we gathered data and how quickly we could gather data from a sequencer. And that could open up a whole new field in infectious diseases diagnostics.
And that's what John Wayne saw as part of the future. And it didn't take him long to convince me that that was probably a great idea. So that's where we started thinking about how we might apply sequencing to the diagnostics of infection. And I had a good background in diagnostics to be able to apply that technology.
What does sequencing make possible that current techniques simply couldn't?
Microbiological culture is slow. It takes at least two days really to go from patient sample to an antimicrobial susceptibility test.
If you are very unwell in hospital, two days is too long to wait to know if you're on the appropriate antibiotics. As antimicrobial resistance rises, as we have more global travel, as people are more immunocompromised in hospitals, we get more unusual bugs and more antimicrobial resistance bugs, and more often the first line therapies just don't work.
So, we need to get the answer to the antimicrobial susceptibility test quicker than two days. In my opinion, microbiologic culture just is not fit for purpose anymore, particularly for severe disease.
What held back metagenomics adoption in a clinical setting?
I worked in the area of diagnostics and rapid diagnostics for a number of years, and I'd applied qPCR to that problem. And that was have starting to have impact back in 2013 with the introduction of sample to answer type machines like the Cepheid gene expert. And just around 2013 was the release of the BioFire Film Array, which was another approach to do multiplex testing in clinical samples. And these technologies, because they were automated from sample to answer, we're starting to have a clinical impact.
But the challenge with those technologies is they were targeted. So, they could not replace culture. They could only be used as an adjunct to culture. They couldn't test for enough antimicrobial resistance and susceptibility. And they only detected the top pathogens for certain diseases and indications.
So, we felt that there was a gap in space in the market for a technology that was as comprehensive or more comprehensive than culture, but that was faster and speed was key. So, we needed to get results from metagenomics as soon as we possibly could.
We started with blood and blood stream infections. And that was a bit naive because that's the hardest challenge. And what we understood after a short while is we did some maths and figured out that you can be septic with one pathogen in a mL of blood. In that same mL of blood, you will have a million leukocytes, the cells within blood that contain DNA.
So that's a million to one ratio straight away. But the challenge here is that there is also another thing to take into account, and that's the size of the genomes. So human genomes are a thousand-fold bigger than bacterial genomes on average. So, if you add those ratios together, you're talking about a billion to one ratio of human DNA to pathogen DNA in a blood sample, you know, at worst.
So that's not good. That's going to take days of sequencing to get a single pathogen read. So, we knew we had to remove human DNA, that's what led us to putting effort into developing host depletion methods.
Part 2 — The Host DNA Depletion Breakthrough
How did serendipity lead to your breakthrough?
So, I took on a PhD student early in my time at the University of East Anglia in my first year in 2013, and his name is Solomon Mwaigwisya.
And Solomon started working on methods for depleting human DNA in blood samples. And that is how we came across ArcticZymes. We were at a (2015) conference in Munich in Germany, the qPCR conference, and Solomon went up to the ArcticZymes booth and he asked; do you have any enzymes that might work in whole blood to deplete or to digest human DNA?
And the person on the booth said, you should try HL-SAN our high salt nuclease. That is heat labile. So, we tried it out and after a while we got some good results, but there's a little bit of a story behind that.
Solomon and I were looking at different ways to deplete host nucleic acid in samples. And we'd come across a few different ways. There was a lysis kit out from Molzym. There was a New England Biolabs kit out that that removed host at the DNA stage. But none of these methods were giving us sufficient host DNA depletion.
We really needed like 10 to the sixfold host depletion. So, 99.9999% removal of human DNA to get that ratio down from a billion to one down to somewhere around a thousand to one, which might be, which might work. So, we weren't achieving that with the commercial methods that we found. So, we went around developing our own and what we did is we, tried different things.
We used differential centrifugation. We used different enzymes for trying to lyse host cells. One of those was phospholipase C, we looked at capturing, human leukocytes using anti CD 45, immuno-magnetic separation. So, we put together a whole load of methods, but one of the steps that we found was weak was that once we had lysed human cells, we couldn't get high activity of nuclease in whole blood.
And so we went looking for nucleases capable of doing that,and we came across the salt activate nuclease from ArcticZymes. And that was really proving to be quite good. But one of the challenges was to get the buffer right. So, we had to try and buffer it so that would be as active as possible in blood.
And so what we did was we tried out a number of buffers. We tried out the buffer that was provided by the company on the website, and that worked and it gave us some host depletion, but it was probably only getting us a hundred, a thousand fold removal of human DNA. And one day Solomon came back from the lab and he was really excited because he'd gotten a kind of a combination of anti CD 45 with some lysis, digestion of cells, and with HL-SAN enzyme to get 10 to the sixth fold depletion of human DNA.
We were all really excited. We went on to combine that enzyme (HL-SAN) with phospholipase C, drop the other parts and come up with a very nice, kind of 45-minute, one hour protocol that was able to get us that 10 to the sixfold depletion. Solomon went on to do some lovely work with that. Then he was towards the end of his PhD, a new postdoc came into the lab, Gemma Kay, and I said, Gemma, you need to get up and running with this metagenomics protocol, it's brilliant.
This host depletion step is where we need to start. And she couldn't get it to work. And we did not know why. And she'd made up the buffer just exactly like Solomon said. And we had to get Solomon back from his write up and we had to try and figure out what was going on. And, so it turned out after a lot of investigation and some reverse engineering, we figured out that Solomon, what he'd written on the tube was what 1X buffer was, but what he'd actually made and what was in the tube was a 10X buffer.
So, we were adding five molar salt, not 0.5 molar salt to the blood. And that had a huge impact on how the enzyme worked and how active it was in blood. So that's how we happened across getting such a highly active and a huge amount of host depletion happening in blood, and that's what we needed for metagenomics to work in an implementable way, so that it would not take too long and not cost too much to do sequencing and use that for infectious diseases diagnostics.
Part 3 — Turning Discovery Into Practice: Trials, Translation and the Push Toward Automation
How can rapid metagenomics move from the lab into standard clinical use?
So,speed is key to everything that we were trying to do with metagenomics. We were trying to get a test, a complicated new technology to work quickly and to provide the data that we needed in a time frame that was faster than culture.
We needed to do all this within a day. We needed to provide same day results to the clinician, and we needed to do that with a sequencing technology. So this hadn't been done before. Now blood wasn't an easy sample to work in. And it was a very challenging ratio of pathogen to human.
So, we moved on, we moved into urine, and then we moved into respiratory tract infection. These are easier sample types. Urine was easy and quite nice to work in. But the economics don't really work out for using metagenomics in urine tract infections. So, you need a clinical syndrome or disease like pneumonia where it's it causes a lot of deaths in hospitals and healthcare systems are willing to pay more money for a test in that particular type of infection.
So, we started working on respiratory tract infections.There are more bacteria in the lung, and when a patient is infected, there's around the same number of human cells. So that ratio isn't as bad. And when you're move that human DNA, you end up with more nucleic acid to work with, and it's just an easier sample type for metagenomics.
So, that application started to really take off. We started working on that as part of a programme called the INHALE trial, where we tried the BioFire Film Array for Pneumonia, the Curetis Unyvero Test and Metagenomics. And we tested all three of those technologies and we chose the one that was most implementable and best at the time, which was the BioFireFilm Array.
But Metagenomics did really well. The problem with metagenomics is it wasn't implementable, right? So, it was too challenging to do in the lab, and I think that's true to this day. So, I'll talk about that a little bit later. But basically, now we had an indication or an application that looked like it was going to work and that was respiratory. And so we needed to get that turnaround time down, into a day.
So, we needed to look at each part of the process and optimize each part of the process. So, the homogenization of the sample to begin with, and moving into a host depletion step. Then moving into DNA extractions, we wanted to keep those times down as much as possible.
We've actually brought the time down for host depletion down to 15 minutes. We do it in one tube, and it's really quick and simple and we still use the same enzyme. And then we moved, um.Uh, we also started using a kit or, or sorry, a library preparation kit,from Oxford Nanopore, which was really fast, which was the Rapid PCR barcoding kit.
And then we were able to add a number of samples, so multiplex samples on the MinION flow cell to costs down. And we were able to reduce the time we needed for sequencing because the Oxford Nanopore flow cells got better and better over time. And then eventually we came out with a protocol that was able to work within about six and a half, seven hours, and that is where we needed to be so we could get a sample in the morning, and we could provide a result to the clinician by the end of the day.
We could give them the pathogens that were present in the sample and some information on antimicrobial resistance. We could follow that up later with some additional AMR data as we got more data off the sequencer. But we certainly had some good information to provide on the same day.
So,we finished the inhale trial and we published a paper from that work, from the metagenomics test that we developed during that time. And one of my PhD students, Charalampous, she was the lead author of that. And, we published that paper in Nature Biotechnology, and that really made a bit of a splash in the field and people were very interested.
Just a year or two later, she finished her PhD and she was hired by Jonathan Edgeworth at Guy's and St Thomas' NHS Foundatiomn Trust (GSTT). And that was just at the start of the pandemic, just like two weeks before the start of the pandemic. I went down to help out at GSTT with some diagnostic development work when the pandemic started.
And so we talked to Jonathan and we decided that it would be a good idea to try to implement metagenomics for respiratory infections of COVID-19 patients while there were in the ICU in St. Thomas's. And we wanted to look for co-infections with bacteria or fungi. And there was a theory at the time that there might be a co-infection etiology to some severe presentations of COVID.
That turned out not to be the case, but the metagenomics testing was very useful anyway, and we published a nice paper on that. Not long after that, I decided to make a move out of academia and into Oxford Nanopore. I'd worked with them for about eight or nine years. At that point, I was very interested in the technology, and I really wanted to drive it forward.
So, I brought some of my techniques with me into Nanopore and developed them further. So, I developed my respiratory metagenomics further within Nanopore. We added the ability to detect viruses and, we made it as robust and as fast as we could. And we released a protocol to the community about a year ago now. And that protocol is being widely used in the community.
So, I guess that brings me up to my recent move in June to the Ellison Institute of Technology in Oxford. And what we're doing at Ellison is that we're going to take metagenomics as an approach and we are going to automate it. And I think that's really the missing step.
And I mentioned earlier that about implementation and why metagenomics is difficult at the moment to implement. And that's really because it's a very long procedure, which is quite laborious. And you have to, maybe there's three or four hours worth of lab work with people in white coats trying to do somewhat complex molecular biology.
We need to take that and simplify it and fix that problem. And, you know, move metagenomics in the same direction as qPCR did, into cartridges, sample to answer. And so that is the challenge that we are going to address at Ellison Institute of Technology. And with my colleague, James Clark we are going to take metagenomics, the approaches that we've developed, we're going to improve them. We're going to automate them. We're going to put them into cartridges in machines that will go from sample to answer. In the first instance, we will certainly go from sample to library and then we will transfer the library onto a MinION flow cell, run the MinION flow cell, analyse the data, provide the answer. And so that work is what we plan to do over the next few years.
And I think that's key if we're going to get metagenomics implemented in hospitals across the world widely, just like we have implemented qPCR widely using sample to answer devices.
