Completion of genome of Aeromonas salmonicida subsp. salmonicida 01-B526 reveals how sequencing technologies can influence sequence quality and result interpretations.

Aeromonas salmonicida subsp. salmonicida is a pathogen that primarily infects salmonids. A strain of this bacterium, 01-B526, has been used in several studies as a reference. The genomic sequence of this strain is available, but comes from pyrosequencing and is the second most fragmented assembly for this bacterium. We generated its closed genome sequence and found a pitfall in result interpretations associated with low-quality genomic sequences.

The bacterium Aeromonas salmonicida subsp. salmonicida is a fish pathogen causing furunculosis [1]. Given its veterinary importance, a sequencing effort was made during the last decade to obtain the genomic sequences of strains of this bacterium. Mainly two technologies were used, pyrosequencing (Roche 454) and sequencing by synthesis (Illumina MiSeq). Pyrosequencing is conducive in generating several errors in the allocation of bases, especially in homopolymeric regions [2]. Even if this technology is no longer available, sequences originating from it are still abundant in databases.

Here, the differences between A. salmonicida subsp. salmonicida genomic sequences resulting from 454 pyrosequencing or Illumina MiSeq were assessed. With the Illumina MiSeq, it is possible to obtain fewer contigs than with 454 pyrosequencing (Fig. 1A). As expected, the N50 values are generally higher for assemblies from Illumina MiSeq than 454 pyrosequencing (Fig. 1B). Only two assemblies, from strains isolated in China [3], do not respect the trend. This may reflect a difference in the assembly tools and parameters used, or even the presence of repeated regions [4], [5].

Fig. 1

Effects of sequencing technologies on genome sequences of Aeromonas salmonicida subsp. salmonicida. Box plots showing distribution of contig size (A) and of N50 values (B) for genome assemblies of A. salmonicida subsp. salmonicida available in GenBank. Information concerning genome sequences used is available in Supplementary Table S1. (C) Clustering and heat map based on average nucleotide identity values. Sequences used are: A. salmonicida subsp. pectinolytica 34melT (assembly: GCA_002735225.1), A. salmonicida subsp. masoucida RFAS1 (assembly: GCA_002313065.1), A. salmonicida A527 (assembly: GCA_002764135.1), A. salmonicida S44 (assembly: GCA_002214305.1), A. salmonicida S68 (assembly: GCA_002214265.1), A. salmonicida S121 (assembly: GCA_002214245.1), A. salmonicida subsp. salmonicida A449 (assembly: GCA_000196395.1), A. salmonicida subsp. salmonicida 01-B526 (454) (assembly: GCA_000234845.2) and A. salmonicida subsp. salmonicida 01-B526 (PacBio) (GenBank: CP027000). 01-B526 (Illumina MiSeq) sequences were only used for comparison and therefore were not deposited.

After the publication of the first complete genome of A. salmonicida subsp. salmonicida, that of the French strain A449 [6], the genome of a virulent strain from Canada was sequenced in order to find biogeographical markers [7]. This made it possible to highlight a genomic island, AsaGEI, whose variants correlate with the geographical locations where the strains were isolated [3], [8], [9]. Since then, strain 01-B526 has been repeatedly used as a reference, both for bioinformatics studies and for wet-lab experiments. Unfortunately, the sequence of this strain is from pyrosequencing data and is the second assembly with the highest number of contigs for this bacterium (Supplementary Table S1).

Given the importance of strain 01-B526 as a reference, we resequenced its genome with Illumina MiSeq as well as with PacBio single-molecule real-time (SMRT) technology, which is known for producing bacterial genome sequences that are often closed [10]. As such, it was possible to obtain the closed chromosome sequence using a combination of Canu version 1.6 [11] and Pilon version 1.22 [12].

To illustrate the biases caused by different sequencing technologies, the three assemblies of strain 01-B526 (454 pyrosequencing, Illumina MiSeq and PacBio SMRT) were compared to other sequences of A. salmonicida found in databases. For this purpose, a clustering was made using average nucleotide identity values (Fig. 1C). As expected, the sequence of strain 01-B526 generated by Illumina MiSeq and PacBio SMRT clustered together, close to the one of strain A449. Surprisingly, the sequence of strain 01-B526 obtained by 454 pyrosequencing was much more distant, even basal to the clade composed of sequences of A. salmonicida strains RFAS1, S44, S68 and S121, which are not of the salmonicida subspecies [13]. The tool used, pyani, aligns fragments of 1020 nucleotides by default for performing the comparisons. Knowing that 77% of the contigs generated from 454 pyrosequencing data were smaller than 1020 nt, this result was not totally unexpected and illustrates the influence of sequencing technologies on subsequent analyses and interpretations.

Accession number

The chromosome sequence of Aeromonas salmonicida subsp. salmonicida 01-B526 has been deposited in GenBank under accession number CP027000.