NEW YORK – Researchers in the UK have used metagenomic nanopore sequencing to rapidly diagnose preterm infants with suspected necrotizing enterocolitis (NEC), a life-threatening gut condition, identifying pathogens and antimicrobial resistance genes in the gut microbiome within a few hours.
The project is among the first clinical applications of nanopore sequencing for infectious disease diagnostics. Other examples include the diagnosis of respiratory infections and of meningitis or encephalitis.
In a paper published in Nature Microbiology on Monday, the team, led by Richard Leggett at the Earlham Institute in Norwich, Lindsay Hall at the Quadram Institute Bioscience in Norwich, and Matt Clark at the Natural History Museum in London, demonstrated using metagenomic shotgun sequencing on the Oxford Nanopore MinIon sequencer and analysis of the results with software they developed to identify gut-associated bacteria and antibiotic resistance genes in a small number of critically ill and healthy infants within five hours. An earlier version of the study was posted as a BioRxiv preprint in 2017 and updated in 2018.
Willem van Schaik, a professor of microbiology and infection at the University of Birmingham who was not involved with the study, said that this application is exciting because NEC is an important cause of death among preterm infants. "[R]apid diagnostics to identify the bacteria (and their profile of antibiotic resistance genes) that contribute to NEC are key to successful treatment," he said in an email. However, widespread clinical implementation of the technology is probably still some time off, he added.
NEC mainly affects very preterm infants and has a mortality rate of about 40 percent. It is currently diagnosed by physical examination, tests like X rays, and routine clinical microbiology testing and is treated initially with broad spectrum antibiotics. "The improvements via nanopore sequencing are the rapid nature and specificity of the test," said Hall, a research leader at the Quadram Institute, via email. "This is very useful with very ill babies that need to be treated as quickly as possible, as well as providing key info for the clinician to treat with antibiotics likely to kill the bacteria," she said. Current tests take much longer and might not catch all types of antibiotic resistance, she added.
According to Leggett, a group leader at the Earlham Institute who has been working with the MinIon technology since 2014, one of the most important aspects of the assay is its speed and the short time to results. He said that conventional clinical microbiology testing, which involves culturing samples on selective media, can take 24 to 48 hours. "With the nanopore approach, we were able to do the same kind of thing in about five hours, and now slightly quicker," he said. This can make a significant difference for treatment, he said: "The sooner you can get this information, the sooner the correct treatment regime can begin."
Besides the nanopore platform, the NanoOK RT software platform for real-time analysis his team has developed has been key, he said. While the sequencing run is ongoing, the software, which is open source and available on GitHub, provides information about bacteria present in the sample as well as antimicrobial resistance genes (AMR), using the Comprehensive Antibiotic Resistance Database (CARD). It also performs a so-called "walkout analysis," using flanking regions from the long nanopore reads to place antibiotic resistance genes in the context of a genome to determine which bacteria they came from.
For their study, the researchers also compared nanopore and Illumina sequencing results for a preterm infant at three different time points and found "very high concordance between the two datasets" with regards to species composition, said Clark, formerly with the Earlham Institute and now a research leader at the Natural History Museum.
The Illumina data picked up a few low-abundance AMR genes that the nanopore data did not but Leggett said those would have been detected if nanopore sequencing had continued for longer.
Initially, he said, his team was concerned the amount of data from the MinIon wasn't enough to capture the true diversity of bacteria in a sample, but that wasn't the case. "By the end of the project, the yields had improved so much that we were able to multiplex samples on a flow cell and get way more data than we needed to," he said. While the first few runs only yielded tens of megabases of data, the lab now obtains tens of gigabases per run.
Accuracy and consistency have also improved significantly over time. When the team first started, the per-base accuracy was only around 70 percent, Clark said, but it has now improved to around 95 percent. The same is true for consistency. "In the early days, consistency was a bit of an issue and you were not necessarily sure what to expect when you loaded certain samples on," Leggett said. "But that consistency has improved enormously."
Identifying the pathogen, which tends to dominate the gut microbiome of a patient, can be very quick with the assay. "That happens almost in seconds once you start sequencing," he said. "And then it takes a little bit longer to get the AMR gene information. But really, in the end, you don't need that many reads to get at that data."
Clark said Illumina sequencing still outperforms nanopore sequencing in terms of per-base accuracy, which is important in cases where antimicrobial resistance is due to a point mutation, for example. "Those are always going to be a bit more challenging for the nanopore data and require you to have a bit more depth in order to reliably call that SNP," he said.
Going forward, the researchers plan to run a multi-center trial to test their assay in a larger cohort of patients treated in the neonatal intensive care unit and compare the results to the standard of care. "With the clinicians, we will review retrospectively how our data would or would not have affected the treatment," Leggett said.
In addition, they plan to develop the method further, speeding it up even more and trying to get additional information out of the nanopore data, which records not only the base sequence but also base modifications. For example, using a rapid library kit has allowed them to shave about an hour from the turnaround time, enabling them to get results in about four hours instead of five.
The metagenomic nanopore sequencing approach might be applicable to other types of patients, too, though the researchers have no plans for testing this yet. "There is a range of conditions related to the gut microbiome and also to conditions that produce leaky gut where pathogens can escape from the gut and cause other problems," Leggett explained.
Van Schaik said there are still a number of challenges to overcome before nanopore sequencing can be used routinely in the clinic. First, nanopore sequencing is still expensive for routine diagnostics. "It could, however, certainly be possible that it finds an application in conditions where [it] provides an important advantage over more traditional, slower diagnostic approaches, like the diagnosis of NEC," he said.
Also, the cost of nanopore sequencing keeps dropping. As part of their study, the researchers ran a sample on a Flongle, an adapter for the MinIon and GridIon that takes $90 disposable flow cells with a lower throughput than regular flow cells. "We got nice results with the Flongle and enough data to perform the analysis from a single Flongle flow cell," Leggett said.
Another impediment to clinical implementation, van Schaik said, is the need for bioinformatics skills to analyze the nanopore data. "Although it appears that the tool presented in this paper is very easy to use, it will still be a challenge to get clinicians to use this approach without support from bioinformaticians," he said. "However, I expect that in the near future, clinicians will become more adept in working with computational data, while the tools themselves will become increasingly easy to use."
Finally, AMR genes identified by DNA sequencing may not always reflect the true antibiotic resistance of a pathogen because the gene may not be expressed or other, unidentified AMR genes may be present. "For this reason, phenotypic antibiotic susceptibility testing of bacterial pathogens will remain the gold standard in clinical diagnostics in the future," van Schaik said. "However, I certainly foresee that rapid nanopore sequencing-based diagnostics could complement traditional phenotypic testing."
For example, nanopore sequencing could generate data within a few hours while phenotypic testing is ongoing, he suggested. "It seems like a no-brainer to act on the basis of nanopore data, even if it is imperfect, while waiting for the outcome of traditional diagnostics (including susceptibility testing), in case there is no other information to guide clinical decision making," he said.