Availability of Nanopore sequences in the genome taxonomy for Vibrionaceae systematics: Rumoiensis clade species as a test case

Introduction

In the last decade, high-throughput DNA sequencing technologies has dramatically expanded the genetic information of all domains of life and viruses, and has contributed to the generation of 2,408 Vibrionaceae genomes deposited in the National Center for Biotechnology Information (NCBI) database (http://www.ncbi.nlm.nih.gov/genome/; December 22, 2017). Whole genome sequencing has become a routine experiment in bacterial taxonomy as the recently emerging “Genome-based Taxonomy” approach provides reproducible, reliable, and highly informative data for phylogenetic inference, differentiating species without the need for specialized skills (Chun et al., 2018). Traditionally, genomic coherence between strains has been determined through DDH experiments, and a cutoff value of 70% DDH for circumscribing species has been recognized as “the golden standard” in prokaryotes taxonomy (Rosselló-Móra & Amann, 2015; Chun et al., 2018). OGRI (Chun & Rainey, 2014) such as ANI (Konstantinidis & Tiedje, 2005; Goris et al., 2007) and genome-to-genome distance (GGD) (Auch et al., 2010) have been utilized to replace traditional DDH, with 95–96% ANI and 70% in silico DDH from GGD corresponding to 70% experimental DDH (Rosselló-Móra & Amann, 2015; Chun et al., 2018). However, whole genome sequencing, particularly methods using Illumina sequencers which are currently the most widely used worldwide, usually require larger facilities and costly maintenance, which could make species identification impractical in smaller laboratories.

Oxford Nanopore Technologies MinION is a palm-sized sequencer, and the ability to produce long reads in real time made Nanopore sequencing an attractive option for genomics. MinION could typically generate 5–10 Gb DNA in a single run (Magi et al., 2017), dramatically reducing the sequencing cost for bacterial genomes. Its accessibility in terms of cost and minimal equipment needs enables rapid acquisition of whole genome data even in non-specialized laboratories. A shortcoming is its relatively high error rate. Nanopore-only assembly, in contrast to hybrid approach combining Illumina reads, still faces challenges regarding its accuracy, which could limit the use of Nanopore sequencing technology (Wick et al., 2017a).

As Vibrionaceae are metabolically and genetically diverse, this group of bacteria have always been at the forefront of bacterial taxonomy being tested for new methodologies (Sawabe et al., 2013). Among 24 clades in Vibrionaceae, the Rumoiensis clade consists of species with diverse ecophysiology, but at the same time the clade is phylogenetically robust (Tanaka et al., 2017). As there are currently only five species in the clade, we were able to finish the genomes of all type strains of the Rumoiensis clade. The aim of this study is to test the availability of Nanopore sequencing in creating a rapid, reliable, and ultimately automatic identification scheme for Vibrionaceae. Using the Rumoiensis clade species as the test case, we compared OGRI and MLSA topology based on data obtained from Nanopore-only assembly and hybrid assembly. Every OGRI determined in this study shows that Nanopore sequencing is applicable for the genome-based taxonomy in Vibrionaceae.

Materials & Methods

Results

To compare the performance of sequencing platforms, the whole genome sequence of V. aphrogenes was obtained using different methods: Nanopore sequencing using MinION (Nanopore), PacBio sequencing using the SMRT system (PacBio), MP and PE reads from MiSeq (MiSeq-MP and MiSeq-PE) (Table S1). Using the Illumina corrected PacBio assembly (PacBio+Illumina) as a reference, overall accuracy and completeness of the assemblies were evaluated. As summarized in Table 1, PacBio or Nanopore alone was able to reconstruct two circular contigs without gaps, each corresponding to the two chromosomes of this bacterium. Furthermore, the Nanopore+Illumina hybrid assembly was highly consistent with the PacBio assembly. Indels derived from tandem repeats with different length (number of repeats) were major differences between the assemblies.

Reference genomes for V. aphrogenes, V. algivorus, V. casei, and V. litoralis were produced by hybrid approach using Unicycler. As Unicycler failed to reconstruct complete circular contigs for V. rumoiensis, Nanopore reads were assembled with Canu, and contigs were manually closed based on the assembly graph. The resulting circular contigs were polished with Illumina reads using Pilon. Using the Unicylcer hybrid or Canu+Pilon assemblies as the references, the error rates of Nanopore-only assemblies were estimated for each species. Consistent with previous reports (Loman, Quick & Simpson, 2015; Wick et al., 2017a), Nanopore-only assemblies had high per base error rates, with 0.4112–1.3644 indels per kb and 0.0265–0.3815 mismatches per kb (Table 2).

Despite the relatively high error rates in Nanopore-only asemblies, taxonomic measures were not highly affected by differences in sequencing or assembly methodologies. Overall, ANI calculated using different assemblies were highly consistent; the comparisons between the same pair of species showed a maximum difference of 0.42% (V. algivorus hybrid and V. casei Nanopore-only: 78.12%, V. algivorus Nanopore-only and V. casei Illumina: 77.70%) (Fig. 1). In silico DDH was more sensitive to the differences in methodology, with values ranging from 79.0 to 100 in the comparison of the same pair of strains (Fig. S1). A considerable difference was also not observed in G+C content, with a maximum difference of 0.04% between hybrid-Nanopore (V. aphrogenes and V. rumoiensis), 0.18% for hybrid-Nanopore and Illumina-Nanopore (V. casei) (Table 3).

To test the compatibility of Nanopore sequencing for species identification with MLSA, eight protein coding genes (gapA, gyrB, ftsZ, mreB, pyrH, recA, rpoA, and topA) used for Vibrionaceae MLSA were retrieved from the Nanopore-only assemblies of fives species and compared with those in the hybrid assemblies. As summarized in Table 4, each gene had zero to two deletions, and occasionally insertions or mismatches, in full length gene comparisons. The error frequencies were reduced if we compare 369–636 bp regions where original eight-gene MLSA used (Tanaka et al., 2017) (Table 4). Error frequencies against Sanger sequences were the same extent of those to Illumina or hybrid assemblies. Using MLSA gene sequences from the hybrid and Nanopore-only assemblies, phylogenetic trees were constructed, and the trees were compared based on their topology. Two trees were topologically identical (symmetric difference of 0) (Fig. S2) with only one node showing differing bootstrap support value.

Discussion

Comprehensive comparative genomics is one of the most promising methodologies in establishing reproducible and reliable criteria toward the next generation microbial taxonomy (Chun et al., 2018). Accelerating use of the genome-based taxonomy increased the demand of fast, high quality, and cost-effective genome sequencing, and the ability to produce long reads in relatively short time, including library preparation, make Nanopore sequencing more attractive. Our data evaluating three major OGRI involving ANI, in silco DDH, and G+C content for the genome-based taxonomy using the phylogenetically robust and genomically distinct Rumoiensis clade species in Vibrionaceae suggests that genome sequences obtained using the ONT MinION are available for genome-based microbial taxonomy.

For Nanopore sequencing of five Vibrio species, the barcoding kit was used to increase the cost-effectiveness. Using two barcodes per species, the sequence data obtained for each species ranged from 0.47 Gb to 2.2 Gb, and the average lengths were between 4.3 kb to 11.0 kb (Table 2). As MinION could typically generate 5–10 Gb DNA in a single run (Magi et al., 2017), the data amounts for each genome obtained in this study are in the range of typically reported values. Combining this with Illumina reads, the Nanopore-based hybrid assembly successfully reconstructed the gap-closed, finished-grade circular genomes involving two major chromosomes commonly possessed by Vibrionaceae species. Ability to reconstruct the complete genomes without the other highly intensive works has great potentials to enhance not only the genome-based taxonomy but also acquisition of the complete genomes for other members of Vibrionaceae. Complete genomes could reinforce our knowledge on the genome plasticity, one of the major topics in elucidating Vibrio biodiversity, pathogenesis, and evolution (Gomez-Gil et al., 2014).

MLSA is a powerful method for inferring the evolutionary history of particular taxonomic groups. In Vibrionaceae systematics, MLSA is particularly important as Vibrionaceae species cannot be identified based on a single molecular marker such as 16S rRNA gene due to the low discriminatory power (Sawabe, Kita-Tsukamoto & Thompson, 2007; Gomez-Gil et al., 2014). As one of the disadvantages of the Nanopore sequencing is the higher error rates compared to Illumina or PacBio sequencing, it is worth evaluating whether these error rates significantly affect the MLSA or not. Unexpectedly, however, low frequencies of mismatch and indel were observed in eight genes typically used for MLSA designed for Vibrionaceae taxonomy (Table 4). In the broad phylogenetic network reconstruction that is generally performed in the initial step of species and/or clade identification of Vibrionaceae, each Rumoiensis clade species forms a robust cluster on each terminal node even if we use gene sequences retrieved from Nanopore-only assemblies. We further conclude that genes from the Nanopore-only assemblies were able to reconstruct the MLSA phylogeny of the Rumoiensis clade species.

Comparison of the Rumoiensis clade species showed Nanopore-only assembly can be utilized to discriminate between species based on OGRI. Additionally, preliminary comparisons of genomes of two strains in the Ponticus clade and three strains in the Splendidus clade of the genus Vibrio sharing 99.9% and 99.8%–99.9% 16S rRNA gene identity, respectively, sequenced using MinION shows 98.7% ANI between Ponticus clade strains and 95.4%–97.0% between Splendidus clade strains, indicating these strains belong to the same species. While this suggests availability of Nanopore sequencing for comparison of closely related strains, caution needs to be taken as high error rates may obstruct the applications such as population genetics and SNP detection. Nonetheless, this fast, reliable and cost-effective means of whole genome sequencing has the potential to advance genome-based taxonomy and development of automated taxonomy solely based on the genomic data.

Conclusions

The complete genomes of five closely-related vibrios were reconstructed using Nanopore sequencing technology. Although Nanopore-only assemblies have previously been described as not being suitable for sequence/allele typing or small variant studies due to high error rates (Magi et al., 2017; Wick et al., 2017a), our dataset shows that Nanopore-only assemblies can be used to discriminate species based on whole genome similarity for taxonomic purposes.

Supplemental Information