Targeting the 16S rRNA Gene for Bacterial Identification in Complex Mixed Samples: Comparative Evaluation of Second (Illumina) and Third (Oxford Nanopore Technologies) Generation Sequencing Technologies

Abstract

Rapid, accurate bacterial identification in biological samples is an important task for microbiology laboratories, for which 16S rRNA gene Sanger sequencing of cultured isolates is frequently used. In contrast, next-generation sequencing does not require intermediate culturing steps and can be directly applied on communities, but its performance has not been extensively evaluated. We present a comparative evaluation of second (Illumina) and third (Oxford Nanopore Technologies (ONT)) generation sequencing technologies for 16S targeted genomics using a well-characterized reference sample. Different 16S gene regions were amplified and sequenced using the Illumina MiSeq, and analyzed with Mothur. Correct classification was variable, depending on the region amplified. Using a majority vote over all regions, most false positives could be eliminated at the genus level but not the species level. Alternatively, the entire 16S gene was amplified and sequenced using the ONT MinION, and analyzed with Mothur, EPI2ME, and GraphMap. Although >99% of reads were correctly classified at the genus level, up to ≈40% were misclassified at the species level. Both technologies, therefore, allow reliable identification of bacterial genera, but can potentially misguide identification of bacterial species, and constitute viable alternatives to Sanger sequencing for rapid analysis of mixed samples without requiring any culturing steps.

A lab in the field: applications of real-time, in situ metagenomic sequencing

“…Although long-reads allowed the classification of 59.41% of the reads down to the species level—no Illumina reads were classified at this level—a significant number of false-positives arose. These results were consistent with observations from [ 61 ], which showed that the bacterial identification at the genus level was reliable. Species-level missclassifications could be partially addressed by employing different—and optimized—bioinformatic approaches for the taxonomic classification [ 45 , 62 ], by sequencing the complete 16S-ITS-23S region of the ribosomal operon [ 63 , 64 ], or by coupling MinION sequencing with complementary quantitative PCR assays [ 60 ].…”

Characterization of Fecal Microbiota with Clinical Specimen Using Long-Read and Short-Read Sequencing Platform

“…The MinION is recently applied as a rapid and cost-effective sequencing platform that generated long reads to allow sequencing of the entire 16S rRNA gene, which may function as an alternative approach to provide high-resolution results regarding bacterial classification [ 20 ]. To evaluate the accuracy of the technology of ONT, the reference standard containing genomic DNAs prepared from eight bacterial and two fungal species were previously subjected to long-read sequencing, followed by two bioinformatics analyses using the CLC genomics workbench or EPI2ME with the SILVA and NCBI 16S reference databases [ 21 ]. The high-similarity taxonomy profiles between the reference standard and analytic results generated using distinct workflows suggested the reliability of ONT results for bacterial classification at the specific level.…”

De novo species identification using 16S rRNA gene nanopore sequencing

“…The current reference-based 16S rRNA gene nanopore sequence analyses are locked to proprietary databases and primers. This may lead to unforeseen biases such as the lack of bifidobacterial detection from mock samples containing several bacterial species ( Kai et al, 2019 ), and a high error rate with respect to species identification ( Winand et al, 2019 ). Reference-based approaches would therefore both fail to identify the Bacteroides species that we identified as being associated with vaginal delivery, and the high levels of Bifidobacterium for the 3-month-old children.…”

“…However, despite several recent advances in nanopore sequencing, the error rates are too high for de novo species identification ( Shin et al, 2016 ). Therefore, all current approaches are based on some kind of reference, or black-box systems for species identification ( Winand et al, 2019 ). This severely limits the widespread application of nanopore sequencing in explorative-based applications.…”