Name: John Rossen
Title: Head of the Molecular and Sequencing Unit of Clinical Bacteriology, University Medical Center, Groningen;
Medical molecular microbiologist, University Medical Center, Groningen
Experience: Chairman, National Working Group on Molecular Diagnostics in Infectious Diseases;
Virologist, Laboratory of Experimental Virology at the Academic Medical Center of the University of Amsterdam;
Medical molecular biologist, St. Elisabeth Ziekenhuis hospital;
Assistant professor of virology, UMC Utrecht
Education: PhD, Utrecht University
Increasingly, public health researchers are evaluating the potential of next-gen sequencing to track and monitor outbreaks. While most of these efforts have been primarily for research purposes or one-off instances of using the technology in an outbreak setting, such as done by the National Institutes of Health last year when it used sequencing to change the management of a Klebsiella pneuomoniae outbreak (CSN 8/22/2012), as the technology continues to prove itself a valuable, cost-effective, and quick means of monitoring outbreaks, it is being integrated into hospitals' and other institutions' protocols for outbreak detection.
The University Medical Center Groningen in the Netherlands has begun integrating whole-genome sequencing of outbreak-causing pathogens in order to design effective screening tools and is also testing a targeted sequencing kit by Pathogenica for identifying pathogens responsible for outbreaks.
John Rossen, who heads the Molecular and Sequencing Unit of Clinical Bacteriology at UMCG, recently spoke with Clinical Sequencing News about his institution's experience with incorporating next-gen sequencing in outbreak detection and management.
How have you incorporated next-gen sequencing for outbreak monitoring?
We started early this year. Once we have an outbreak situation and begin screening [for the outbreak strain], we also need to screen lots of patients that are not in the hospital anymore. They send samples by mail, so we get lots of samples, sometimes between 600 and 1,000 samples.
When you screen that [many samples] with conventional methods you find all sorts of bystander bacteria, which are interesting to know, but not important for the outbreak management. What we do now is as soon as we have the outbreak and the strain that caused the outbreak, we do next-gen sequencing, we do whole-genome sequencing, and we try to find a unique SNP or a unique part in the genome of that specific outbreak strain to design a screening assay. Most of the time we use real-time PCR for that. So we go in for a unique marker in the genome of the outbreak strain that is not present in similar strains, so when we screen the patients, we only find this specific outbreak and not all kinds of bystanders.
Do you also use next-gen sequencing to identify the outbreak strain?
For the identification of the outbreak, in the case of bacteria, we use conventional methods, but we also now have a pilot study in which we use the Pathogenica HAI kit to see what we can find, to see which kind of strain is present and which resistance genes it has. That's more like a screening assay. It's not whole-genome sequencing, but is more focused on specific targets.
The other thing is, when we really have bacteria that we cannot identify [with conventional methods], we can do deep sequencing to see if we can find what is present.
Can you share the results of the Pathogenica pilot?
Right now we're in the middle of it, so I can't tell you that many details on the results yet. I can tell you that we are now performing the Pathogenica kit in an outbreak screen to see whether the results we find fit with results we find with more conventional methods. So far it looks fine, but I can't give you more details yet.
What is the problem with hospital outbreaks — are they becoming more common and how are they typically managed?
There are more bacteria that more or less start to contaminate each other, infect each other. In the hospital we see this occurring more and more and it's very important to keep these outbreaks as small as possible because we have had problems already, where hospitals have to close the wards, and that's of course not something that you want. Not only for economical reasons but also related to taking care of patients.
Closing the ward is the fastest way to get rid of the outbreak — we're quite successful in that. But the people that work on the ward don't like it very much. For them it's a disaster for patient care and economically.
What are the conventional methods you use to identify outbreak strains?
For bacteria, most of the time we culture it. We do have many real-time PCR assays to detect both viruses and bacteria. In the case of viruses, that is enough. You can just identify what virus it is and do some sequencing to get an idea on the virus type. But in the case of bacteria, you also want to know the resistance pattern, and we still do that phenotypically. We do conventional culture, plating the sample on culture plates and then we also identify the phenotypic resistance using Biomerieux's Vitek machine and we also use a microarray [from the Netherlands-based Check-Points] to detect which genes are involved in the resistance.
What are the advantages of using a sequencing-based test, like Pathogenica's kit versus traditional methods to identify the outbreak strain?
It's quicker, indeed. That's one of the very important reasons to do the pilot. We're screening the intensive care unit two times a week and sometimes samples arrive on Thursday and when you finally get the results [using conventional culture techniques] it might be six days later. With Pathogenica, we are able to get the results in a shorter time.
We're still working on making the workflow more efficient because you have to do some library prepping and DNA isolation and that adds to the time required. So we're working on optimizing this workflow, but the Pathogenica run itself is very fast.
Another advantage of the kit is that you don't need any bioinformatics. There is plug-in software, which tells you what kinds of pathogens are present and if they contain resistance genes.
Currently, [one drawback with the kit is] that if we have a mixed sample, we cannot correlate the resistance gene to a specific pathogen. If you detect three different bacteria and one resistance gene, you don't know to which of the three pathogens the resistance gene belongs. So you need to do another assay afterwards. But at least you know that there is a resistance gene present in the bugs you detect.
But the convenience of having this software that gives you a report, that states the bugs and the resistance genes you found, without doing any difficult analysis, I think makes the kit something that can be applied in laboratories where there might not be a bioinformatics department or in smaller hospitals.
You mentioned you are also doing whole-genome sequencing of the bacteria species once it's identified to figure out a marker around which to design an RT-PCR assay to test patients. What is the protocol for that?
As soon as we have an outbreak and the strain, we immediately sequence the whole genome and assemble it using CLC Bio software. To find unique SNPs, we have a homebrew pipeline. For next-gen sequencing, we now have [Illumina's] MiSeq [for whole-genome sequencing].
The Pathogenica kit in Europe is [run on the] Ion Torrent PGM. So actually, at the moment, we have both machines in the [microbiology] lab … and we can also make use of equipment in the [pathology lab]. They have an Ion Torrent, and the clinical genetics department has HiSeqs and MiSeqs.
Can you walk me through an example of how you used sequencing to track and monitor an outbreak, and how that potentially changed the outcome?
It starts with monitoring. We have a person that wrote a software program to monitor all kinds of bacteria that are identified at a certain ward. As soon as we have more than two of the same [pathogens] in a short period of time, there's an alarm that makes us realize an outbreak could be going on.
Then we have the bug isolated from the first patient so we can immediately start to isolate the DNA. That can be done in a few hours. And then for whole-genome sequencing, we are using the Nextera XT kits from Illumina. So, from when we have the colony, we can in one day do library prep. Then we run the sequencer. We use 2x250 base paired-end reads on the MiSeq. The machine runs for 42 hours. Then we have our results. We can do the assembly quite efficiently.
Then we look in the database where we have all kinds of data stored of whole-genome sequences of different bugs and we compare and see if we can find a unique marker. And then we can start screening with that unique marker. Within one week we can develop a real-time PCR test and then we start screening. We think that the process will be shortened and within a couple of years it will be available after three to four days.
What if the bacteria strain is not culturable?
We don't detect it. There are some difficult [bacteria] that when out of the body, they are not culturable anymore. In those cases, it would be nice to try the next-generation sequencing technique. That is the next step. For all the things that we cannot detect anything, we want to use deep sequencing to see whether we can find the bacteria.
Currently we've tried BioFire's FilmArray, [which is based on multiplexed PCR] but not all the pathogens are present in the assay.
What have been the major challenges of using next-gen sequencing for outbreak monitoring?
The analysis and bioinformatics, which are getting better now because there's more and more software. In the case of typing, one of the programs we are currently using is SeqSpere, a program designed by Ridom, [a German company that develops software for DNA sequencing analysis].
Once we have assembled the whole genome, we can put it into the software and compare the different genomes. That's a nice program that you can use in a very easy way to determine the core genome and you can compare every gene within the core genome. And you can compare between strains.
That's helpful, but still bioinformatics is a challenge. It's not only that you need the technicians, but you need bioinformaticians and ideally you'd like to have bioinformaticians that also have knowledge of microbiology. That's one challenge.
I think automation is another of the challenges, which people are working on. For example, to make a library for the Ion Torrent [system] is quite a hassle, but I know that they're working on a machine called Ion Chef to automate the different steps.
Automation is extremely important, not only because you save hands-on time, but it also increases quality. Once it is automated you can do barcode tracing, and the chances that samples get mixed up or that you make pipetting errors are decreased.
Do you think next-gen sequencing techniques will be adopted throughout hospitals for outbreak monitoring?
I think it will start in the university hospitals, and kits like Pathogenica's, which do part of the analysis will make it also interesting for smaller hospitals. Already in the Netherlands there is a lot of interest from non-university hospitals and several are already collaborating with us to see whether it could fit in their outbreak management. So I think there will be a future for next-gen in hospitals for outbreak management.