Skip to main content
Premium Trial:

Request an Annual Quote

Oxford Nanopore Sequencing Shows Promise for Diagnosing Prosthetic Joint Infections


NEW YORK (GenomeWeb) – Clinical researchers from the UK are developing a method for diagnosing infections associated with prosthetic joint implants that promises to provide results in a matter of hours as opposed to a week or more.

The work, which leverages Oxford Nanopore sequencing technologies, was presented at the European Congress of Clinical Microbiology and Infectious Diseases in Amsterdam earlier this week, one of several presentations that discussed the potential of the technology for rapid infectious disease diagnosis in near-patient settings.

Teresa Street, a research scientist with the Nuffield Department of Medicine and a lecturer in genetics at the Institute of Human Sciences at the University of Oxford, discussed her group's attempts to validate the technology as part of a completely culture-free method for diagnosing prosthetic joint implant infections.

To illustrate the current problem, Street cited data collected by the UK's National Joint Registry on joint replacement surgeries in 2017. Of approximately 113,000 total knee replacement surgeries performed that year, 6,500, or about 6 percent, were revision surgeries, and nearly a quarter of those surgeries took place because of an infection or suspicion of infection. Elbow replacement surgeries showed similar rates of infection-related revision surgeries, while shoulder and hip replacements were around 17 percent and 16 percent, respectively.

Street said that the current gold standard is bacterial culture from periprosthetic tissue samples collected during surgery, although another option is culturing from sonicated explanted prostheses in saline.

She noted, however, that culture from tissue samples is relatively insensitive, with detection rates around 65 percent, and is "a very busy process with many steps." Her team wanted to know whether it could improve the detection of orthopedic device infections by doing it faster, in culture-negative cases, and directly from clinical samples.

In theory, metagenomic sequencing can do all these things. It essentially involves extracting all DNA or RNA directly from a clinical sample, sequencing it, and comparing it to a reference genome to identify pathogens.

Street and colleagues demonstrated proof of principle for the approach using Illumina MiSeq sequencing in a study published in 2017 in the Journal of Clinical Microbiology. The workflow they devised involved removing a prosthetic device during surgery then placing it in saline and sonicating it to obtain approximately 40 mL of sonication fluid — essentially " the largest volume we can easily handle in the lab, which is allowing us to maximize the number of cells we can extract DNA from," Street said at ECCMID.

The sonication step potentially increases the number of bacterial cells available in the sample because it disrupts the bacterial biofilm. The DNA is extracted from the sonicated samples, cleaned, and prepared into libraries, which are then sequenced.

Although they achieved relatively high levels of species-level sensitivity and specificity (88 percent for both), they learned a few things for future studies, Street said.

"We learned that human DNA is really problematic," she said at ECCMID. "In 97 percent of our samples, more than 90 percent of the sequence data that we generated was human, so we are obviously just throwing that away."

They also learned that it is important to generate metagenomic profiles of culture-negative samples for comparisons, because contamination is a problem. This contamination can arise from multiple sources including sample collection, sample processing in the lab, and from the actual kits and reagents used in sample processing.

Most importantly, they learned that they needed more sequencing reads in order to be able to do more than just identify species, which is why the group has since switched to nanopore sequencing. Essentially, the longer read lengths and additional data generated by nanopore sequencing enables higher read coverage of reference genomes.

"It's exciting," Street said. "It's portable. You can generate longer read lengths, but crucially you can analyze your data in real time. … It's not really a stretch to think that it's something you could use next door to surgery during the timeframe of surgery."

In a follow-up email, Street noted that her team has had to change a few things to its protocol for nanopore sequencing, most importantly adding a much more efficient human DNA depletion step. "The nanopore library prep is allowing us to incorporate a PCR amplification step, which is giving us longer fragments to carry forward into the sequencing and more DNA to actually sequence," she said.

The group is also employing a bioinformatics pipeline developed by the University of Oxford's Nick Sanderson called CRuMPIT (clinical real-time metagenomics pathogen identification test). Key components of this pipeline include a program called Centrifuge, which is doing the taxonomic classification, and Minimap2, which is mapping reads to reference genomes. The CRuMPIT pipeline was described in a BMC Genomics paper published last year.

In an unpublished proof-of-principle study, Street's team analyzed DNA samples left over from the previous Illumina study and has thus far found excellent concordance. In a few cases, they were also able to identify two separate species from a previous genus-level identification (e.g. Bacillus cereus and B. thuringiensis instead of Bacillus spp.).

Street also noted that the time to generate the reads necessary for species identification has been drastically reduced using nanopore sequencing.

The group is now thinking about how to further optimize their method for nanopore sequencing. Specifically, they are testing a detergent-based sample prep that selectively lyses human cells, followed by an endonuclease that degrades the human DNA before moving to the bacterial lysis step.

They are also testing Oxford Nanopore's rapid PCR barcoding kit which allows for 10 nanograms of DNA input to counteract the fact that they are depleting most of their viable DNA during the human DNA depletion step. This kit also employs a six-minute PCR elongation step which creates longer fragments for analysis, and more PCR cycles to increase yield. Furthermore, it enables multiplexing, which reduces the overall cost of sequencing.

Thus far, Street said, their analyses using this improved protocol have a high degree of concordance with culture testing, and in fact they have been able to detect positives for certain species that were culture-negative. In addition, in a few Staphylococcus-positive cases so far they have been able to identify antimicrobial resistance genes.

In one sample, they were able to detect two different Staphylococcus organisms, one of which they could identify (S. haemolyticus) and one of which they could not. They later used MALDI-ToF mass spec to identify this organism as S. caprae, and realized they couldn't initially detect it because it was not in their reference database, underscoring the fact that metagenomic sequencing is only as good as the reference database being used.

Street's team is not the first to use metagenomic sequencing to identify infectious organisms in joint replacements. A team led by Robin Patel at the Mayo Clinic developed its own shotgun metagenomics method using Illumina sequencing, also obtaining promising results and recognizing some of the same challenges as Street noted — namely, the difficulty in reducing the amount of human DNA in a sample and maximizing bacterial DNA reads.

Patel's team also noted that data analysis is time-consuming and requires specialized bioinformatics expertise, both of which are barriers to routine implementation. Earlier this year, Patel and colleagues published a paper in the Journal of Clinical Microbiology evaluating a bioinformatics package from Rockville, Maryland-based firm CosmosID for this purpose, noting that it has the potential to provide "fast, reliable bacterial detection and identification from metagenomic shotgun sequencing data derived from sonicate fluid for the diagnosis of prosthetic joint infections."Street noted that despite the early promising results of her and others' research, "there is still much work that needs to be done before this method could be considered for implementation. We haven't sequenced enough samples with our new methods on [Oxford] Nanopore yet to be able to address in a more statistical way the data regarding identifying background [and] contaminating reads. We need more data before we can begin to work out cutoffs or thresholds in our read numbers so we can be more confident that what we are identifying represents true infection."