Multilocus sequence analysis reveals different lineages of Pseudomonas anguilliseptica associated with disease in farmed lumpfish (Cyclopterus lumpus L.)

The bacterium Pseudomonas anguilliseptica has in recent years emerged as a serious threat to production of lumpfish in Norway. Little is known about the population structure of this bacterium despite its association with disease in a wide range of different fish species throughout the world. The phylogenetic relationships between 53 isolates, primarily derived from diseased lumpfish, but including a number of reference strains from diverse geographical origins and fish species, were reconstructed by Multi-Locus Sequence Analysis (MLSA) using nine housekeeping genes (rpoB, atpD, gyrB, rpoD, ileS, aroE, carA, glnS and recA). MLSA revealed a high degree of relatedness between the studied isolates, altough the seven genotypes identified formed three main phylogenetic lineages. While four genotypes were identified amongst Norwegian lumpfish isolates, a single genotype dominated, irrespective of geographic origin. This suggests the existence of a dominant genotype associated with disease in production of lumpfish in Norwegian aquaculture. Elucidation of the population structure of the bacterium has provided valuable information for potential future vaccine development.


Introduction
Introduction and widespread use of cleaner fish, such as lumpfish (Cyclopterus lumpus), for control of sea lice infestations (Lepeoptheirus salmonis and Caligus elongatus) in Atlantic salmon (Salmo salar) aquaculture in Norway presents novel virologic, parasitic and bacteriological issues. Since the start of production of farmed lumpfish, the bacteria most commonly associated with disease in this fish species have been Vibrio anguillarum [1],'atypical' Aeromonas salmonicida [2], Moritella viscosa, Tenacibaculum spp., and Pasteurella sp. [3,4]. Following its first isolation in 2013, Pseudomonas anguilliseptica has however emerged as an additional and serious threat to lumpfish production in Norway, and the number of outbreaks has since increased annually [5]. P. anguilliseptica associated mortalities have also been reported in lumpfish farmed in Scotland [6], Ireland, and the Faroe Islands (Scholz; Christiansen; unpublished data). P. anguilliseptica is recognized as an opportunistic pathogen primarily affecting fish in marine and brackish environments. It was originally described in 1971 as the causative agent of "red spot disease" (sekiten-byo) in pond cultured Japanese eel (Anguilla japonica). Isolated in 1981 from European eel (Anguilla anguilla) in Scotland [7], the bacterium became a frequently occurring pathogen associated with eel farming throughout Europe. Initially, P. anguilliseptica was considered to be exclusively associated with disease in eel culture [8]. However, the bacterium has since been recognized to display little or no host specificity as it has been shown to be pathogenic in a range of cultured and wild fish species in different geographic areas (Table 1). With the exception of a small number of isolates from farmed wolffish (Anarhichadidae) and a single isolation from seawater farmed rainbow trout (Oncorhynchus mykiss) in 2018 [9], P. anguilliseptica infection appears to be particularly associated with disease in lumpfish in Norway.
Little is known of the population structure of this bacterium despite its association with disease in numerous fish species over a wide geographic range. While phenotypic studies [12,13,21], identified few differences, Random Amplification of Polymorphic DNA (RAPD) fingerprinting separated the population into two genetic clades, with one clade almost exclusively represented by eel isolates.
A subsequent serological study [22] identified two serologically distinct groups, termed serotypes O1 and O2, consistent with the two RAPD groups identified previously. While serotype O1 were isolated from a diverse range of fish species, serotype O2 isolates were almost exclusively associated with eels. The concordance observed between the genetic and serologic characteristics culminated in the proposal of two separate clonal lineages for the bacterium [23].
Given the genetic and serological diversity demonstrated within the species and existence of both species-specific variants and broad host range variants, the aim of the present study was to elucidate the phylogenetic relationships amongst Norwegian isolates of P. anguilliseptica isolated from farmed lumpfish. A multilocus sequence analysis (MLSA) scheme comprising nine housekeeping (HK) genes was therefore designed to phylogenetically reconstruct the population structure of a variety of host derived, geographically disparate and historically distinct isolates of P. anguilliseptica. The knowledge generated may shed light on infection dynamics, allow development of specific diagnostics and identify strains relevant for future vaccine development.

Confirmation of isolate identity
A total of 53 isolates were included in the analysis. These were sampled from a range of different spatiotemporal origins and fish species ( Table 2). All isolates originated from clinically infected fish and were cultivated on blood agar w/2% NaCl at 15˚C for 48 hours. Matrix Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry (MALDI-TOF) was performed on all isolates, while 16S rRNA gene sequencing and/or phenotypic characterization was used in some cases to positively confirm species identity prior to the MLSA.

MALDI-TOF.
MALDI-TOF analysis was performed on a Biotyper Microflex LT (Bruker Daltonics, Bremen, Germany). In-house main spectra (MSP) were generated for P. anguilliseptica isolate NVIB 50-2084 and type strain NCIMB 1949. For MSP generation, proteins from each reference strain were extracted according to the manufacturer's protocol (S1 Protocol).
Following incorporation of constructed MSPs into the reference library, single colonies of putative P. anguilliseptica were subjected to MALDI-TOF analysis by the Direct Transfer Method according to the manufacturer's instructions. Briefly, single colonies were smeared with a toothpick as a thin film onto two successive spots on a MALDI target plate. After drying at room temperature, the spots were overlaid with HCCA matrix solution (saturated α-cyano-4-hyoxycinnamic acid in 50% acetonitrile 2,5% trifluoracetic acid) and air-dried. Identification was then performed using the standard Biotyper Database supplemented with the in house generated MSPs. Similarity scores �2.0 in relation to any particular MSP were considered to represent good identification to the species/genomovar level.
Phenotypic characterization. Phenotypic characterization was conducted on a representative selection of six isolates derived from various fish host species and geographic-/temporal origins (NVIO 10973, NVIO 8905, NVIO 11214, NVIB 50-2255, NVIB 50-1910 and NVIO 9976; Table 2). The characteristics examined were: colony morphology, gram staining, motility, anaerobic respiration and hemolysis on blood agar. Biochemical properties such as oxidative/fermentative (O/F) production of acid from glucose, as well as the ability to produce decarboxylases from the amino acids arginine, lysine and ornithine (ALO), were also tested. Production of oxidase was tested using the DrySlide Oxidase (BD, Franklin Lakes, USA).

DNA extraction, PCR and sequencing
Template DNA was isolated using the High Pure PCR Template Preparation Kit (Roche Applied Science, Penzberg, Germany) in accordance with the manufacturer's protocol. Concentration and purity of nucleic acid in the final eluate was assessed with the NanoDrop 2000 Spectrophotometer (ThermoFisher Scientific, Waltham, USA). All samples were diluted to 2 ng/μl before being stored at -20˚C.
The following housekeeping (HK) genes were included in the MLSA assay: rpoB (RNA polymerase beta subunit), atpD (ATP synthase F1 beta subunit), gyrB (DNA gyrase beta Table 1. Species of wild caught and cultured fish shown to develop disease due to infections with P. anguilliseptica.
Cycle sequencing of purified PCR products was based on a reaction mixture containing (per reaction) 0.5 μl BigDye v3.1, 2.0 μl 5X Sequencing Buffer, 2.0 μl M13 primer (2.5 μM), 2.0 μl purified PCR product and 3.5 μl MilliQ-water, and involved 96˚C for 1 min, 25 cycles of 96˚C for 10 seconds, 50˚C for 5 seconds and 60˚C for 4 min, followed by cooling to 4˚C indefinitely. This was performed using an Applied Biosystems1 Veriti1 96-Well Thermo Cycler (ThermoFisher Scientific, Waltham, USA). Sequence products were purified using the Big-Dye1 XTerminator Purification Kit (ThermoFisher Scientific, Waltham, USA) in accordance with its protocol. Capillary electrophoresis of purified sequence products was performed using the 3500XL Genetic Analyzer (ThermoFisher Scientific, Waltham, USA) according to the manufactures protocol.

Data analysis
Forward and reverse strands for each amplicon were sequenced for all isolates and imported to Geneious 11.1.5 (Biomatters, Auckland, New Zealand) for contig assembly and further processing. Contigs were manually checked and edited and re-sequencing performed when necessary. Contigs were then trimmed at the 5´and 3´-ends to exclude primer sequences and areas of poorer quality towards the amplicon ends. All sequences were trimmed to preserve the reading frame. Final processed sequences were imported and concatenated in Microsoft Excel. Concatenated sequences were then re-imported to Geneious to construct the final alignment. Complete record of partial gene sequences for all isolates was submitted to GenBank (accession nos. MW684870-MW685347).

Multi-Locus sequence analysis
Based on the concatenated alignment, a maximum likelihood phylogenetic reconstruction was run in MEGA X using the TN93 substitution model [25] (identified as optimal by the model test module). Number of bootstraps was set to 1000, while all other inputs were set to default. The phylogenies were similarly reconstructed for individual genes and for a concatenated amino acid alignment (LG substitution model [26] used for the latter).

Confirmation of isolate identity
MALDI-TOF analysis of samples submitted for the study positively identified all 53 isolates as P. anguilliseptica with similarity scores �2.0. 16S rRNA gene sequencing of 39 isolates also demonstrated concordance with MALDI-TOF results.
With the exception of production of arginine dihydrolase (isolate NVIO 8905 negative), the studied isolates were phenotypically similar. The results of phenotypical testing are listed in Table 4.

Multi-Locus sequence analysis
The concatenated alignment of nine housekeeping genes (5,364bp) displayed average sequence identities of 99.9% between all isolates. Phylogenies inferred from MLSA based on nucleotide sequences revealed seven genotypes distributed amongst three major lineages (Fig 1). Inspection of the tree shows that lumpfish-isolates are dispersed thoughout depending on country of origin (i.e. Norwgian isolates in lineages 1 and 2, Faroese in lineage 1 and Irsish in lineages 1 and 3), while isolates from other fish species (irrespective of geography) belong almost exclusively to lineage 1. The type strain NCIMB 1949 from Japanese eel in Japan constitutes one exception to this, as it forms a singleton genotype within lineage 3. Phylogenetic reconstruction based on concatenated amino acid sequences displayed compatible topologies (Fig 2). Maximum likelihood trees generated from single gene alignments displaying genetic heterogeneity (rpoD, rpoB, carA, and atpD) are attached as (S1-S4 Figs). For aroE, glnS, ileS, recA and gyrB, complete genetic homogeneity was observed across all 53 isolates.

Discussion
The present work represents the most comprehensive phylogenetic study to date of the fish pathogenic bacterium P. anguilliseptica originating from a range of different geographic localities, fish species and time of isolation.
MLSA revealed a high degree of genetic similarity irrespective of geographic origin, fish host species or time of isolation. The significant degree of genetic conservation identified amongst isolates retrieved from a wide geographic area over a period of 21 years suggest 'purifying' selection pressure towards a restricted fish parasitic/pathogenic niche. Evolution of virulent host associated lineages from ancestral environmental strains has previously been described from Francisella tularensis [27]. As in previous studies relating to P. anguilliseptica, all isolates studied here originated from clinically infected fish. No environmental isolates were available. Had such isolates been available they may have revealed a greater intraspecific diversity. While the possibility that fish pathogenic P. anguilliseptica represent expansion of a fish pathogenic clone against a background of environmental, possibly non-virulent strains cannot be ruled out, there is, however, no literature published to date suggesting the existence of environmental strains of this bacterium.
For dihydrolase/decarboxylase testing + denotes a strong purple reaction while-denotes a weak or lack of colour change. https://doi.org/10.1371/journal.pone.0259725.t004 A degree of phylogenetic heterogeneity was, however identified. The phylogenetic patterns identified appear to be more related to geographic origin than chronology or host fish species. The Norwegian isolates, originating with the exception of two isolates from wolffish and ballan wrssse, from farmed lumpfish, were distributed between two of the three major lineages (lineage 1 and 2, Fig 1). Lineage 1 is the most numerous of the three, with isolates originating from most geographic localities and fish species. Approximately a third of all Norwegian isolates are situated within this lineage, while the rest are situated in the exclusively Norwegian lineage 2. Although the Norwegian isolates were distributed amongst two phylogenetic lineages, no distinct differences were registered between these lineages in terms of domestic geography or time of isolation. MLSA based on concatenated amino acid sequences displayed a pattern consistent with nucleotide sequence analysis in the formation of three major lineages.
Following identification of the various lineages by MLSA, MALDI-TOF MSP's were generated from NVIB 50-2084 and NCIMB 1949, representing MLSA lineages 1 and 3 respectively (Fig 1). Subsequent MALDI-TOF analysis of the entire study collection utilizing these MSP's revealed that MALDI-TOF may be capable, with further refinement, of distinguishing between the different lineages observed, and may therefore prove a valuable future diagnostic tool for discrimination of these lineages. The results of phenotypic characterization were in compliance with previously published descriptions [12,13,21] with the exception of a single isolate which was negative for production of arginine dihydrolase (Table 4). A complementary analysis of all isolates included in the present study showed three isolates (NVIO 8827, NVIO 8905, and NVIB 50-1579), all situated within lineage 2, negative for arginine dihydrolase production. This appears to be a variable trait only within this lineage. The existence of phenotypic variation beyond the phylogenetic resolution achieved through MLSA can thus not be ruled out.
Previous studies have suggested the existence of antigenic variation within P. anguilliseptica. Seven of the isolates studied in López-Romalde et al. [15,22] were used in the present study (see Table 1). All included isolates representing serotype O1 were situated within the same phylogenetic lineage (Fig 1) and it is reasonable to assume that all isolates within this lineage probably belong to serotype O1. Notably, all but one of the Irish lumpfish isolates included in this study clustered closely with the type strain NCIMB 1949, isolated from Japanese eel in Japan, within lineage 3 (Fig 1). The serological differences identified by López Romalde et al. [22] between representatives of lineage 1 (O1) and NCIMB 1949 (O2) suggest that these Irish isolates might potentially also belong to serotype O2. Future antigenic characterization would be required to determine whether this is the case. Irish isolate NVIO 11302 deviates, however, from this pattern and is situated within the geographically spread lineage 1 (Fig 1). This situation could either be due to the existence of several genotypes associated with disease in Irish lumpfish or possibly indicate spread of infection via import of Norwegian lumpfish-roe to Ireland.
Of the Norwegian isolates, 8 of 26 were situated within lineage 1, likely representing serotype O1. The remaining, and thereby the majority of Norwegian isolates, were situated in lineage 2, which shows comparable phylogenetic separation from lineage 1 as are at least as phylogenetically distant, if not slightly more distant to lineage 1 than NCIMB 1949 (serotype O2) (Fig 1). This raises the question as to whether lineage 2 may represent a previously yet undescribed serotype. The fact that all Norwegian isolates are distributed between two of the major phylogenetic lineages (1 and 2), possibly representing different serotypes, complicates identification of a suitable candidate strain for vaccine development. Characterization of the serological properties of Norwegian isolates will therefore be necessary with respect to the potential development of a mono-or multi-genotype vaccine for future use in lumpfish.
Interestingly, isolates of identical genotype were identified from Norwegian lumpfish and diseased Atlantic salmon in Finland (Fig 1). The identical genotype and the pathogenic nature of the bacterium [12] suggests that P. anguilliseptica might potentially pose a threat to cohabiting salmonids in commercial net pens in Norway. However, considering the ever-increasing number of outbreaks among lumpfish and the lack of reported infections in cohabiting salmon, this indicates that Norwegian salmon are not highly susceptible to P. anguilliseptica infection. While future whole genome sequencing studies may reveal sub-MLSA genotype differences related to host specificity, establishment of a challenge model to test salmon susceptibility to the Norwegian lumpfish strain should be considered to evaluate the potential risk of cross-species transmission.

Conclusion
The nine-locus MLSA employed in this study identified a considerable degree of genetic similarity among isolates of P. anguilliseptica, separating the studied isolates amongst seven closely related genotypes. Several genotypes were identified amongst Norwegian lumpfish isolates, although the majority belonged a single overarching lineage. Antigenic study of representative isolates from each genotype should be evaluated as a basis for development of vaccines against this important lumpfish pathogen.