Discussion
Culture-based techniques have been the standard for detecting pathological microorganisms in clinical specimens. Metagenomic sequencing has emerged as an alternative methodology for overcoming the drawbacks of conventional bacterial cultures. 16S rRNA gene amplicon sequencing is useful for identifying causative bacteria from clinical samples with a relatively low bacterial load, such as ophthalmological specimens.
Metagenomic analysis with NGS platforms has been reported to identify Achromobacter in a case of scleral buckle infection, Corynebacterium propinquum, in contact lens-related microbial keratitis,14 15 and can also be applied for detecting unculturable bacterial pathogens. In a clinical context, this technique is useful, especially for identifying causative agents that are difficult to grow in culture, such as nutritionally variant streptococci, even after antibiotic administration.28 29 However, such short-read NGS technology has some disadvantages in terms of speed and accuracy, which leads to low taxonomic resolution and time consumption for bacterial identification. The portable nanopore sequencer MinION used in this study can analyse long DNA sequences covering the full-length bacterial 16S rRNA gene. Our previous study showed that full-length 16S rRNA gene amplicon sequencing with MinION provides a higher taxonomic resolution than short-read sequencing using the short-read NGS platform.26 Accumulating evidence demonstrates promising results with MinION sequencing, which enables bacterial identification at the species level.8 Another intriguing feature of MinION sequencing is that it generates sequencing reads in real-time, allowing for a far shorter turnaround time for data processing. In our approach, bacterial identification of clinical specimens could be completed within a total analysis time of approximately 4 hours in the shortest time (figure 2). This would offer a significant advantage for the rapid diagnosis of infectious diseases.
Figure 2Schema of the platform of 16S rRNA gene amplification analysis with MinION. It takes at least 4 hours from sample collection to the sequencing.
Given the excellent discriminatory power of MinION long-read sequencing, we validated its efficacy in identifying pathogenic bacteria in a wide range of ophthalmological samples. MinION sequencing successfully identified bacterial species from the vitreous fluid, corneal scrapes and abscess samples. The profiles of dominant taxa analysed by MinION sequencing were consistent with those obtained from the bacterial cultures in all samples tested. As for Moraxella, species-level identification is not attainable by culture-based methods, and mass spectrometry has been utilised for accurate taxonomic classification. We successfully identified M. nonliquefaciens by MinION sequencing, which was confirmed by MALDI-TOF MS analysis. In case 6, both fungi and P. aeruginosa were detected in the culture. Fungi were suspected to be the primary causative organisms, which may explain the low abundance of bacteria estimated from the 16S rRNA gene sequencing results. Although our experimental platform is limited to identifying bacteria, nanopore amplicon sequencing also allows the identification of pathogenic fungi by using universal primer sets that target a genetic marker, such as the internal transcribed spacer of the fungal rRNA gene.
These results suggest that full-length 16S rRNA gene amplicon sequencing via MinION is a reliable and practical tool for the rapid identification of pathogenic bacteria with high resolution in ocular specimens. All the identified organisms matched the culture, smear, or clinical findings in each case. Thus, we have made the decision of causative organisms from these various kinds of analyses.
Low et al reported 16S rRNA gene analysis using a portable nanopore sequencer in two cases of keratitis.30 Jun et al also used a nanopore sequencer in eight cases of endophthalmitis.31 Low et al reported the advantages of gene analysis using MinIONin detecting causative organisms after administering antimicrobial eye drops. Jun et al described the advantage of gene analysis using MinION for the early detection of endophthalmitis in the anterior aqueous humour at the first visit. In both reports, the number of cases was small, and the authors emphasised the need for a greater number of cases.
Low et al stated that the collection of keratitis specimens is affected by the material and shape of the swabs used. In this study, we used a direct spatula to scrape the ulcerative tissue, providing sufficient DNA for analysis. The amounts of the specimens from conjunctival abscesses and endophthalmitis were small, that is, 0.1 mL of leaking purulent material and 2.5 mL of aspirated vitreous. However, we were able to detect the causative organism in both cases. The primer sequence for the bacteria used by Jun et al is unclear, but the kit manual they used indicates that some types of bacteria may not be detectable. In contrast, we used a primer set optimised to amplify the 16S rRNA genes across a broad range of bacterial species, the specificity of which has been confirmed in the analysis of the intestinal microflora.26 We used the database provided by Oxford (EPI2ME Fastq16S), as in Jun et al.
Genetic analysis using MinION has some limitations. First, compared with conventional methods such as MiSeq, MinION might be less accurate in determining individual bases and have a higher error rate. However, this can be compensated by long-chain decoding. Second, as was seen in cases 5 and 6, it is often the case that a variety of bacteria are detected. Therefore, the possibility of contamination by indigenous microorganisms should be carefully considered. In a review of diagnostic methods for ocular infections, Eguchi also stated that the diversity of micro-organisms detected by genetic analysis requires careful analysis to determine whether they are contaminated or indigenous microorganisms.32 Due to the high sensitivity of amplicon sequencing, even a small amount of DNA contaminated during the sample preparation can lead to false-positive results. For reliable detection of pathogenic bacteria, it is required to minimise the risk of DNA contamination both in collecting clinical specimens and the subsequent experimental procedures. Furthermore, common background species should be carefully distinguished from infectious pathogens. Especially in a low-level infection or a sample with low read counts, the presence of the contaminants makes it difficult to interpret the sequencing results. Although it is challenging to completely rule out the influence of these contaminants, incorporating negative controls processed precisely the same way should be a minimum requirement to ensure the validity of the results and distinguish contaminants from true pathogens. Third, taxonomic resolution is only at the species level and not at the strain level. Therefore, virulence factors cannot be identified, and genetic analysis cannot differentiate dead micro-organisms from live ones.
Even if these factors are considered, MinION could provide high resolution and rapid identification from a small specimen amount, and we believe that it will contribute greatly to daily clinical practice.
