Description of New and Amended Clades of the Genus Photobacterium

Phylogenetic relationships between species in the genus Photobacterium have been poorly studied despite pathogenic and ecological relevance of some of its members. This is the first phylogenetic study that includes new species of Photobacterium (validated or not) that have not been included in any of the previously described clades, using 16S rRNA sequences and multilocus sequence analysis (MLSA) in concatenated sequences of gyrB, gapA, topA, ftsZ and mreB housekeeping genes. Sequence analysis has been implemented using Maximum-parsimony (MP), Neighbour-joining (NJ) and Maximum likelihood (ML) treeing methods and the predicted evolutionary relationship between the Photobacterium clades was established on the basis of bootstrap values of >75% for 16S rRNA sequences and MLSA. We have grouped 22 species of the genus Photobacterium into the following 5 clades: Phosphoreum (comprises P. aquimaris, “P. carnosum,” P. iliopiscarium, P. kishitanii, P. phosphoreum, “P. piscicola” and “P. toruni”); clade Profundum (composed of P. aestuarii, P. alginatilyticum, P. frigidiphilum, P. indicum, P. jeanii, P. lipolyticum, “P. marinum,” and P. profundum); clade Damselae (two subspecies of P. damselae, damselae and piscicida); and two new clades: clade Ganghwense (includes P. aphoticum, P. aquae, P. galatheae, P. ganghwense, P. halotolerans, P. panuliri and P. proteolyticum); and clade Leiognathi (composed by P. angustum, P. leiognathi subsp. leiognathi and “P. leiognathi subsp. mandapamensis”). Two additional clades, Rosenbergii and Swingsii, were formed using a phylogenetic method based on 16S rRNA gene, although they are not confirmed by any MLSA methods. Only P. aplysiae could not be included in none of the established clade, constituting an orphan clade.


Introduction
The family Vibrionaceae (Gammaproteobacteria) is a diverse group of Gram-negative bacteria that includes following genera: Vibrio, Photobacterium, Aliivibrio, Catenococcus, Echinomonas, Enterovibrio, Grimontia and Salinivibrio [1,2]. According to divergence in the 16S rRNA gene sequence and phenotypic characteristics, the genus Photobacterium nowadays comprises more than 28 validated species [3], which are widespread both in the marine environment (seawater, sediments and marine animals) and in saline lakes [4]. Some species are bioluminescent and form specific bioluminescent mutualisms with marine fish [5]. In addition, several strains have been reported to be pathogenic for both poikilothermic and homeothermic animals and are capable of producing important disease outbreaks with a high economic impact [4,6].
Photobacterium species display varied phenotypic, physiological and ecological characteristics, although all of them are chemoorganotrophs, possess Q-8 as the predominant respiratory lipoquinone and present C 16:1 and C 16:0 as their major fatty acids. Recently, this genus has been revised on the

Strains and Culture Conditions
A total of 53 Photobacterium strains, including type strains and environmental isolates were analysed. Fourteen strains presumptively belonging to the genus Photobacterium were isolated from captive fish with signs of disease [38,39]. Strains were routinely cultured on Tryptic soy agar or broth (TSA or TSB) (Difco) supplemented with 1.5% (w/v) NaCl (TSAs or TSBs, respectively) and incubated at 22 • C for 2 to 5 d. Stock cultures were stored at −80 • C in TSBs with 15% (v/v) glycerol.

DNA Extraction and PCR Amplification
Bacterial genomic DNA was extracted according to methodology described previously [38]. Six genes were studied: gyrB (DNA gyrase B subunit), gapA (glyceraldehyde-3-phosphate dehydrogenase A), topA (DNA topoisomerase I), ftsZ (GTP-binding tubulin-like cell division protein), mreB (cell wall structural complex MreBCD) and 16S rRNA. PCR primers used for amplification and sequencing of these genes are listed in Supplementary Table S1

Phylogenetic Data Analysis
The sequences of 16S rRNA (n = 53) and the gyrB, gapA, topA, ftsZ, mreB genes (n = 47) for MLSA of all strains of Photobacterium tested were aligned using MUSCLE software [41]. For MLSA analysis individual genes were aligned and then concatenated using FaBox software [42]. Sequence sizes (nucleotide number) of each of the five housekeeping genes analysed in this study were 2421, 774, 2653, 1143 and 1044 nt for gyrB, gapA, topA, ftsZ, mreB genes, respectively.
Recombination events in sequence alignments were analysed using the Recombination Detection Program Beta v4.95 (RDP4) [43], following the methods previously described [44,45]. Phylogenetic analysis was performed using the program MEGA6 [46]. Maximum-parsimony (MP) analysis was performed for all gene fragments and the tree was obtained using the Subtree-Pruning-Regrafting (SPR) algorithm [47]. For comparative purpose, phylogenetic analyses using Neighbour-joining (NJ) and Maximum likelihood (ML) treeing methods were carried out using, in both cases, the Jukes-Cantor model [48]. In all cases, gaps and missing data treatment was accomplished using complete deletion strategy. Bootstrap analyses were performed using 1000 replications and a bootstrap of ≥75% was used to provide the confidence estimation for clades in the phylogenetic tree.

Phenotypic Characterization
Phenotypic characterization of the 18 Photobacterium strains isolated from diseased fish is shown in Supplementary Table S3. According to the biochemical and physiological profiles, 15 strains belong to P. damselae subsp. damselae and 3 strains were classified as P. damselae subsp. piscicida. The intraspecific variation among two subspecies was obtained for the following features: motility; nitrate reduction; acids from: sucrose, melibiose, D-cellobiose, maltose and D-trehalose; and for the utilization as unique carbon and energy source of the following substrates: glycogen, tween-40, N-acetyl-D-galactosamine, β-methyl-D-glucoside, D-raffinose, D-sorbitol, succinic acid, D-L-lactic acid, bromosuccinic acid, succinamic acid, L-alanyl glycine, L-asparagine, L-aspartic acid, L-glutamic acid, L-serine, glycerol and α-D-glucose 1-phosphate. On the other hand, the intraspecific variation among P. damselae subsp. damselae strains was recorded for the following characteristics: lysine decarboxylase, acetoine production, amylase, gelatinase, lipase and haemolysin production; whilst, for P. damselae subsp. piscicida strains the intraspecific variation was obtained for acetoine production, amylase and acids from L-arabinose (Supplementary Table S3). The profiles of API 20E for P. damselae subsp. damselae were 2014144, 2015144, 6014144 and 6015144, while for P. damselae subsp. piscicida were 2004024, 2004025 and 2005025. In the case of API 20NE, the profiles for P. damselae subsp. damselae strains were 5342334 and 5342344 and for P. damselae subsp. piscicida strains the profile was unique 4142344. As it can be seen, the phenotypic patterns are very variable among the strains of the same subspecies and therefore, they are not adequate to apply for evolutionary or phylogenetic studies.

Phylogenetic Studies of the Photobacterium Genus
The results of the phylogenetic analysis performed on the 16S rRNA gene sequences are shown in Figure 1 and Supplementary Figures S1 and S2. All strains identified as either P. damselae subsp. damselae or P. damselae subsp. piscicida cluster in a single, tightly packed group with 97.0% bootstrap support for MP, with 95.0% for NJ and ML treeing methods (Figure 1, Supplementary Figures S1 and  S2, respectively). No internal boundaries appeared between both subspecies, since their sequences display almost no variation and form a tight monophyletic branch constituting a homogeneous group. For a more robust phylogenetic analysis, a MLSA approach using set of five housekeeping genes was performed, which have been proven to be useful for taxonomic and phylogenetic studies of the Vibrionaceae family [20,[49][50][51][52]. MLSA using MP grouped 17 from 18 strains of P. damselae with a bootstrap value of 99% (Figure 2), constituting the clade Damselae. This result is confirmed by the use of NJ and ML treeing methods, including all the 18 strains of P. damselae, both with a bootstrap value of 94% ( Supplementary Figures S3 and S4).
The position of other Photobacterium species studied on the basis of 16S rRNA and MLSA was always external to the Damselae clade, constituting five additional clades that include all the other 21 Photobacterium from 30 described species (Figure 2). Clade Ganghwense, a new proposed clade, includes six species: P. aphoticum, P. aquae, P. galatheae, P. ganghwense, P. halotolerans and P. proteolyticum on the basis of the MLSA approach at a bootstrap value of 86% (Figure 2). In addition, P. panuliri may be included in this clade on the 16S rRNA gene sequence at bootstrap values of 88, 86 and 88% for MP, NJ and ML treeing methods (Figure 1, Supplementary Figures S1 and S2). This clade grouped species that possessed <5% GC (mol %) (48.6 ± 3.4%) [15,16,23,28,29]. This result is consistent with that obtained by Lucena et al. [23], who reported that P. aphoticum presented a close relationship with P. halotolerans and P. ganghwense and with the results of Liu et al. [28] who found that P. aquae and P. aphoticum had higher 96% 16S rRNA sequence similarity. On the other hand, Rivas et al. [16] found that P. halotolerans presented a close relationship with P. ganghwense and Gomez-Gil et al. [25] established a clade formed by P. ganghwense, P. halotolerans, P. aphoticum, P. panuliri and P. aquae but P. galatheae formed an orphan clade. However, the later species was closely related to P. halotolerans according to the results of Machado et al. [32].
Clade Leiognathi, is a new proposed clade, consists of two subspecies: P. leiognathi subsp. leiognathi and P. leiognathi subsp. mandapamensis and P. angustum according to the MLSA at a bootstrap value of 75% ( Figure 2). This result is similar to those reported by other authors [5,18].
The inclusion of P. frigidiphilum into the clade Profundum (16S rRNA gene sequence analysis) or into the clade Phosphoreum (MLSA analysis) is uncertain, although this species has been related to P. indicum (clade Profundum) by several authors [15][16][17]52]. The unexpected close proximity of P. frigidiphilum to P. kishitanii in the concatenated tree (Figure 2), which is due to the higher similarity shared in the housekeeping genes (98.0% gyrB, 99.3% mreB, 99.5% topA, 100% gapA and 100% ftsZ) than observed for the 16S rRNA gene sequences (97.1%). In future studies with P. frigidiphilum should be considered a reassessment of the sequences available.  On the basis of the 16S rRNA gene sequence analysis, four species: "P. atrarenae," P. gaetbulicola, P. lutimaris and P. rosenbergii clustered on the basis of 16S rRNA gene sequences at a bootstrap of 82% (Figure 1) constituting the clade Rosenbergii. However, applying the MLSA approach these species and P. sanctipauli formed paraphyletic branches and they could not be included in a clade (Figure 2 and Supplementary Figures S3 and S4), except in the case of the use of NJ methods that grouped P. lutimaris and P. rosenbergii in a cluster with a bootstrap value of 96% (Supplementary Figure S3), result similar to that reported by other authors [20]. These species shared >20% DNA-DNA hybridization (from 21.5% to 22%) and possessed <5% GC (mol %) (49.7 ± 3.9%) [17,21,24,49]. Results that are partially similar to those reported by several authors [24,52], who reported that the highest degree of similarity of P. atrarenae was with P. rosenbergii and P. gaetbulicola.
The species: P. sanguinicancri and P. swingsii on the basis of exclusively of 16S rRNA gene sequence possess a bootstrap value of 80% (Figure 1), presenting <5% GC (mol %) (45.2 ± 2.26%) [25,33], which could constitute a new clade named Swingsii. Phylogenetic analysis of 16S rRNA gene sequence revealed that P. sanguinicancri is closely related to P. swingsii [33]. However, this inclusion of these two species in a clade in the MLSA approaches has been accomplished at bootstrap values <75% (Figure 2, Supplementary Figures S3 and S4) and this not confirmed the proposed of this clade.
Only the inclusion of P. aplysiae is uncertain and constitutes the unique orphan clade. Several authors, on the basis of 16S rRNA gene sequence, revealed that P. aplysiae was closely related to P. alginatilyticum [35,37] and to P. swingsii [25], although these results are not based on MLSA analyses, and, therefore, it is necessary to include more strains in further phylogenetic studies to elucidate the inclusion of P. aplysiae in any clade.

Intra-and Interspecies Nucleotide Sequences Variation
The mean intra-and interspecies nucleotide sequence similarities for the different genes tested are shown in Table 2. The intraspecies gene similarities in the Damselae clade are high with six genes tested (>83%), regarding the geographical location and isolation host. For the other clades established in the present study, the intraspecies gene similarities were variable, with values between 44.0% and 99.9% for clade Phosphoreum; between 79.4% and 98.4% for clade Profundum; between 12.0% and 97.8% for clade Ganghwense; and between 15.8% and 99.2% for clade Leiognathi.

Conclusions
The MLSA has showed to be a robust technique suitable for elucidating phylogenetic relationships among P. damselae strains and between P. damselae and other species of the Photobacterium genus and hence, its use is necessary for taxonomy of this microbial group. By using this assay 22 from 31 of the described until now species of Photobacterium have been adequately discriminated and 5 clades have been proposed on the basis of MLSA approach. Two additional clades, Rosenbergii and Swingsii, were formed using the 16S rRNA gene as phylogenetic approach, although they are not confirmed by any MLSA methods. Thus, only P. aplysiae is not included in any cluster and it constitutes an orphan clade. All the new recently described species (validated or not) have also been clustered in the defined and proposed clades.