BALTIMORE – As nanopore sequencing is becoming widely used, researchers at the US Food and Drug Administration are exploring the use of Oxford Nanopore Technologies' MinIon device as a tool for foodborne pathogen surveillance.
In a webinar hosted by the FDA last month, Jason Neal-McKinney, a research microbiologist in the agency’s Applied Technology Branch at the Pacific Northwest Food and Feed Laboratory, demonstrated the utility of MinIon sequencing for identifying, characterizing, and tracking foodborne pathogens when combined with Illumina MiSeq data to achieve hybrid genome assemblies.
According to Neal-McKinney, whole-genome sequencing is “widely used” by researchers and analysts at the FDA’s Office of Regulatory Affairs (ORA) laboratories to conduct outbreak and strain tracking as well as to characterize an organism’s gene content to identify potential determinants of virulence and antimicrobial resistance mechanisms.
As part of the agency’s routine tasks, he said, ORA labs constantly sequence bacterial isolates from collected samples and submit their genomes to GenomeTrakr, a genome database hosted by the National Center for Biotechnology Information (NCBI) that is primarily used for foodborne pathogen source tracking and outbreak detection.
So far, most of the genomes submitted to GenomeTrakr by the FDA were sequenced using the Illumina MiSeq, which Neal-McKinney called “the workhorse of the ORA labs.” While MiSeq can produce high-quality sequencing data with fairly large throughput, the short reads can result in fragmented de novo genome assemblies, he said, making some genetic features inaccessible to researchers.
Meanwhile, the MinIon is capable of generating long-read sequencing reads, which can be used to construct complete genomes. But the relatively low accuracy of the technology often means the assembled genomes contain errors, making downstream data analysis more challenging, Neal-McKinney said.
To paint a more complete picture of pathogen genomes, his group therefore sought to combine MinIon long-read and MiSeq short-read sequencing data, using a bioinformatic pipeline called Unicycler.
The researchers tested the method on Campylobacter jejuni — including two reference isolates with complete reference genomes and two experimental isolates from their archive — and compared the accuracy of the hybrid assembly with that of genomes constructed using MiSeq and MinIon data alone.
“We chose Campylobacter in particular because it has a relatively small genome, but the genome has a lot of repeats and low GC content that can make assembly difficult,” Neal-McKinney said.
In brief, DNA from the C. jejuni isolates was extracted using the Qiagen DNase Blood and Tissue kit. While the MiSeq libraries were generated using the Illumina DNA Prep kit, MinIon libraries were prepared with the 1D2 library prep kit. The respective libraries were sequenced on the MiSeq using the 2 X 250 cycle chemistry or on the MinIon using the FLO-Min107 flow cell.
The group then assembled the genomes using MiSeq data only with the Spades pipeline, Canu for MinIon sequences, and Unicycler for the hybrid assembly.
Echoing the advantages and limitations of each platform, MiSeq generated high-quality data but fragmented assemblies due to the short read lengths, while genomes created using solely MinIon data contained errors that can hinder gene characterization. Meanwhile, the combination of short and long read data in Unicycler was able to create complete circularized genomes and reveal extrachromosomal sequences, such as the plasmids in C. jejuni.
The team next applied the hybrid assembly method to five major foodborne pathogens: Escherichia coli, Salmonella enterica, Listeria monocytogenes, C. jejuni, and Vibrio parahaemolyticus. In addition to sequencing the pathogen samples on MiSeq and MinIon, Neal-McKinney also included data from the Illumina iSeq 100 to test how the performance of the assembly methods may vary within different short-read sequencing platforms.
The results showed that both iSeq and MiSeq data can be combined with MinIon data to achieve more accurate genome assemblies. However, genomes assembled using 250 base pair MiSeq chemistry contain fewer contigs than assemblies using 150 base pair iSeq data, he said.
Similar to the previous data, the results also showed that the hybrid assembly method was superior in revealing the chromosomal and extrachromosomal elements of each contig.
Because iSeq and MinIon have lower throughput than MiSeq, Neal-McKinney said these platforms may be more suited to sequence sporadic real-time isolates. His lab typically analyzes 12 samples in each MiSeq run, and if there are not enough samples, archived isolates from the freezer are added to fill the space.
Lastly, his team looked more closely at the L. monocytogenes isolates that were collected by FDA investigators from environmental swabs at the same firm in 2016 and 2021. Previously, the isolates were sequenced using MiSeq and submitted to GenomeTrakr, and the short-read genome assembly allowed the researchers to analyze single-nucleotide variants, which revealed the isolates were closely related.
To supplement the short-read data, the group also sequenced the isolates on MinIon and generated hybrid genome assemblies. These revealed additional genetic features, including extrachromosomal elements and genetic variations such as inversions, insertions, and deletions.
“What’s really exciting is that this existing MiSeq data that we have in the GenomeTrakr database can be supplemented with MinIon sequencing to further investigate strains that we’re interested in and generate a more complete picture,” Neal-McKinney added.
Moving forward, his team plans to test new nanopore sequencing chemistry and products as they’re released. However, the rapidly evolving nature of the technology can also be “a double-edged sword,” Neal-McKinney said. “It’s nice that there are all these new kits and technologies available, but by the time you get one really established, they tend to move on to the next product release.”
In an email to GenomeWeb after the webinar, an FDA spokesperson said the agency is currently testing Oxford Nanopore’s new Q20+ chemistry, which promises to boost sequencing quality, though the researchers do not yet have any results to share yet. They anticipate the quality of MinIon reads will continue to improve with the new chemistry, further improving the quality of the assemblies.
In addition, Neal-McKinney said his team is keen to test new assembly and analysis methods. Although a powerful tool, Unicycler starts with a short-read assembly and uses the MinIon data to settle discrepancies and bridge the gaps within the genome, he said. Therefore, the outcome of Unicycler can be impacted by the quality of the short-read assembly. He therefore plans to try out other hybrid genome assembly methods that use long-read data first.
Furthermore, he plans to try real-time MinIon sequencing using GPU-accelerated base calling. One of the advantages of MinIon over MiSeq is that the data are generated in real time and bases can be called using a GPU-enhanced computer, he said, whereas MiSeq sequencing data won’t be available until the end of a run, which takes nearly two days for 2 X 250bp sequencing.
This also means the nanopore sequencing technology is well suited for rapid detection and confirmation of isolates, according to the FDA spokesperson. Although the agency is interested in developing nanopore sequencing as a rapid field test, it is currently only using it as a research device, the spokesperson noted, adding that a field test would also require a pre-enrichment step to increase the technology's sensitivity for foodborne pathogens in environmental samples.
In addition, even if MinIon sequencing were to be deployed in the field, library prep would still require the use of either a thermal cycler or calibrated heat blocks, and sequencing with the MinIon would also require a computer or an advanced MinIon model with an integrated computer, the spokesperson pointed out.
In terms of cost, the biggest difference between the MiSeq and MinIon is in the startup costs, according to the spokesperson, with a MinIon starter pack being less expensive than a MiSeq. As for reagent costs, MinIon and MiSeq can vary significantly based on the volume ordered and negotiated contracts, she said, but MiSeq would likely be more cost effective for large numbers of samples. Meanwhile, MinIon may be more cost effective when there are fewer isolates to run, although FDA scientists are currently assessing which application is best suited for the MinIon in terms of cost, the spokesperson said.
Neal-McKinney said he intends to apply the hybrid genome approach to analyze metagenomic or mixed samples, since the method is capable of assigning the sequences to specific DNA molecules and genomes.
Despite the utility of MinIon sequencing, the FDA spokesperson stressed that nanopore sequencing is not the only long-reading sequencing technology adopted by the agency. “The FDA has been a longtime user of Pacific Biosciences long-read sequencing and has owned several models including a recent Sequel IIe with HiFi sequencing capacity,” she said. The FDA uses PacBio sequences to close bacterial genomes and their mobile elements to assist in outbreak investigations and infectious disease control, the spokesperson said, and one strength of the technology is that the quality of its data eliminates the need for hybrid assembly.
Additionally, while Neal-McKinney’s research group at the FDA’s Pacific Northwest Laboratory is not currently working on PacBio or synthetic long-read sequencing, FDA scientists in other laboratories are evaluating these technologies, she said.
Beyond that, the FDA is also evaluating other new sequencing technologies, including the Illumina NextSeq 2000, Pacific Biosciences Sequel IIe, and the Element Biosciences Aviti, to assess their cost-effectiveness to improve food safety and public health. “This includes assessments of new equipment, chemistries, and software,” the spokesperson said.