Chryseobacterium potabilaquae sp. nov., Chryseobacterium aquaeductus sp. nov. and Chryseobacterium fistulae sp. nov., from drinking water systems

Abstract

A polyphasic taxonomic study was conducted on three strains isolated from drinking water systems that had previously been deposited as Chryseobacterium species at the Spanish Type Culture Collection in order to complete their classification. Strains CECT 9293^T, CECT 9390^T and CECT 9393^T were isolated from sites in Barcelona, Spain, in the framework of a project aimed at generating the first MALDI-TOF database specific for bacteria present in water for human consumption. Their partial 16S rRNA sequences showed that their closest relatives among the type strains of Chryseobacterium exhibited 98 % similarity or less, supporting their taxonomic novelty. At the same time, comparison between them revealed that strains CECT 9293^T and CECT 9393^T could perhaps be related at the species level as they shared 99.5 % similarity. However, whole genome sequencing was performed and the subsequent calculation of relatedness indices, average nucleotide identity and estimated DNA–DNA hybridization, ruled out that possibility and confirmed instead that each of the strains should be considered a separate species in the genus Chryseobacterium . Having clarified their status, we also performed phylogenomic analyses and searched for possible environmental or non-type material sequences that could be related to any of them at the species level. In parallel, the strains were characterized phenotypically and compared to their closest relatives to determine diagnostic traits to support their formal proposal. The proposed species are Chryseobacterium potabilaquae sp. nov. with the type strain CECT 9293^T (=LMG 32084^T), Chryseobacterium aquaeductus sp. nov. with the type strain CECT 9390^T (=LMG 32085^T) and Chryseobacterium fistulae sp. nov. with the type strain CECT 9393^T (=LMG 32086^T).

The genus Chryseobacterium , formerly part of the family Flavobacteriaceae [1, 2], contains more than 120 species registered at LPNS (https://lpsn.dsmz.de/genus/chryseobacterium accessed on May 2021) [3]. Its taxonomy is in a status of flux and, apart from various previous emendations, the recent major changes include its assignment to the newly defined family Weeksellaceae [4] and the transfer of 24 species to the neighbouring genera Epilithonimonas (nine), Halpernia (three), Kaistella (11) and Planobacterium (one) [5]. Chryseobacterium species are widely distributed in soils, plants, animals and man-made environments and many of them are related to aquatic habitats [2]. In particular, they have been detected as part of the culturable bacterial fraction of drinking water systems in different countries [6–8].

Quality and health regulations of tap water in developed countries require microbial monitoring to guarantee the absence of pathogens, based on bacterial faecal indicators and heterotrophic bacteria enumeration. Heterotrophic plate counts are important for various reasons including monitoring bacterial regrowth and biofilm formation, the inclusion of opportunistic pathogens such as Aeromonas , Klebsiella and Pseudomonas , which may have consequences for public health, and the interference with the enumeration of faecal indicator micro-organisms. The characterization of microbial populations in drinking water is also important for water management, as it can reveal impaired functioning of treatments and/or distribution pipelines [8].

During a survey of drinking water bacteria populations carried out in the framework of a project aimed at generating the first MALDI-TOF database specific for bacteria present in water for human consumption, several bacteria were isolated from tap water and MALDI-TOF was used to resolve the identity of the isolates obtained. The most abundant genera in the drinking water system were Acidovorax and Blastomonas , in terms of both the total number and their presence across samples. Other important genera were Acinetobacter , Aquabacterium , Bosea , Brevundimonas , Caulobacter , Citrobacter , Polaromonas , Pseudomonas , Rhizobium and Variovorax [8]. Three isolates that were not identified through a combination of MALDI-TOF and 16S rRNA gene partial sequencing were further investigated and were found to be related to Chryseobacterium , although they do not correspond to any described species. This study was undertaken to resolve the taxonomic position of the three strains through a characterization comprising phenotypic, phylogenetic and genomic aspects of their biology.

Isolation and maintenance

The three strains of Chryseobacterium characterized in this study, CECT 9293^T, CECT 9390^T and CECT 9393^T, were obtained from the Colección Española de Cultivos Tipo (CECT). Strain CECT 9293^T had been isolated from a drinking water treatment plant in Sant Joan Despí (Barcelona, Spain; GPS: 41° 21′ 9.5″ N 2° 3′ 00.5″ E) on 20 August 2013; while strains CECT 9390^T and CECT 9393^T had been isolated from a drinking water distribution network in Castelldefels (Barcelona, Spain; GPS: 41° 16′ 9.4″ N 1° 59′ 41.8″ E) on 15 May 2017. All three strains were obtained from membrane filters on water ISO plate count agar (WiPCA), incubated at 22 °C for 3 days, streaked on the same medium until pure cultures were obtained, submitted to MALDI-TOF MS bacterial identification procedure and reported as non-identified. After deposition in CECT, they were further stored in lyophilized format for a long-term maintenance, using meso-inositol (5 %) as cryoprotectant. To ensure they would be available on a second culture collection in a different country, they were deposited at LMG (Laboratorium voor Microbiologie, Universiteit Gent, Belgium) where they are kept as LMG 32084^T, LMG 32085^T and LMG 32086^T, respectively.

16S rRNA gene phylogeny

A partial 16S rRNA sequence was obtained by PCR amplification and Sanger sequencing [9] and subsequently compared by blast searches that placed all three strains as unnamed Chryseobacterium species.

Complete sequences of 16S rRNA genes were later obtained from the sequenced genomes and, after confirming they matched (100 % coincidence) the respective partial sequences, they were compared through blast [10] and EzBioCloud [11] tools with their closer neighbours. Maximum-likelihood (ML; Fig. 1) and maximum-parsimony (MP) trees were inferred from 16S rRNA gene sequences under the GTR+CAT model by the GGDC web server [12] available at http://ggdc.dsmz.de/ using the DSMZ phylogenomics pipeline [13] adapted to single genes.

Fig. 1.

ML tree inferred from 16S rRNA gene sequences under the ML analysis under the GTR+CAT model and rooted by midpoint-rooting. The branches are scaled in terms of the expected number of substitutions per site. Phylogenies were inferred by the GGDC web server [12] available at http://ggdc.dsmz.de/ using the DSMZ phylogenomics pipeline [13] adapted to single genes. The numbers above the branches are support values when larger than 60 % from ML (left) and MP (right) bootstrapping.

Similarities between the 16S rRNA gene sequence of strain CECT 9390^T and the type strains of its closest neighbours were 98.0 % to Chryseobacterium glaciei , 97.8 % to Chryseobacterium polytrichastri and 97.5 % to Chryseobacterium gambrini . Strain CECT 9293^T shares a 96.9 % sequence identity to Chryseobacterium daecheongense , 96.9 % to Chryseobacterium hispalense and a 96.7 % to Chryseobacterium candidae . Strain CECT 9393^T, on the other hand, shows 97.2 % identity to C. hispalense , 97.2 % to C. candidae and 97.1 % to C. daecheongense . As can be inferred, the 16S rRNA sequence from strains CECT 9293^T and CECT 9393^T are highly similar, namely 99.5 % identical, since they relate to the same organisms. Similarity values lower than 98.5 % for this gene are generally considered as indicative of a separate species status, while above the threshold more resolutive methods are needed [12, 14]. Thus, while the taxonomic novelty of strain CECT 9390^T can be presumed by these results, in the case of strains CECT 9293^T and CECT 9393^T it needs to be elucidated if they represent one novel species together or one novel species each.

The position of the three strains within the Chryseobacterium phylogenetic tree, as defined on the basis of the 16S rRNA gene, is shown in Fig. 1. While strain CECT 9390^T merges with C. glaciei , the pair CECT 9293^T and CECT 9393^T forms a separate branch, without a particular link to other species. All three strains are located on the core of the genus Chryseobacterium , as recently defined by Nicholson et al. [5].

Genome analysis and phylogeny

Genomic DNA was isolated using Jena Bioscience (Diffractia) following the standard protocol recommended by the manufacturer. The integrity of the extracted DNA was checked by visualization in a 2.0 % (w/v) agarose gel electrophoresis. Its purity and quantity were checked by measuring the absorbance at 260 and 280 nm with a spectrophotometer Nanodrop 2000c (Thermo Scientific) and calculating the ratio A260/A280. Genome sequencing of the three strains was achieved at Central Support Service for Experimental Research (SCSIE) of the University of Valencia (Valencia, Spain). In the case of strains CECT 9293^T and CECT 9393^T, this was done using Illumina Miseq technology with 2×250 paired-end reads. The reads were analysed for quality control using FastQC, a tool developed by Babraham Bioinformatics to check raw sequencing data. After filtering, the remaining reads were assembled using SPAdes 3.9.0 software [15]. A plot, coverage versus length of the contigs, was performed to help in the choice of the parameters for contig filtering. After the filtration of contigs (500 bp and 10–50×kmer coverage), evaluation of the final assembly against a reference genome was done with the software Quast version 4.3 [16]. In the case of strain CECT 9390^T, genome sequencing was achieved using Sequel PacBio RS II technology (SMRT Link version 7.0) and assembled with the Hierarchical Genome Assembly Process (HGAP4) de novo assembly analysis application.

The bioinformatic tool CheckM version 1.0.7 [17] was used to assess the genome quality prior to annotation using Prokka version 1.12 [18] and rast version 2.0 [19]. The process of quality assessment of reads, read-processing, assembly and annotation with Prokka was carried out in Linux OS, other tools were accessed online. The minimal standards for the quality of genome sequences and how they can be applied for taxonomic purposes [14] have been observed in this study.

Similarity between genomes was established using in silico DNA–DNA hybridization (isDDH) with the Genome-to-Genome Distance Calculator (GGDC 2.1) [12], average nucleotide identity (ANI) [20] with JSpecies software (http://jspecies.ribohost.com/jspeciesws/), and average amino acid identity (AAI) [21] with AAI matrix tools (http://enve-omics.ce.gatech.edu/g-matrix/). Phylogenomic analysis was performed with UBCG (Up-to-date Bacterial Core Gene) [22]. This software tool is available for download at EzBioCloud [11] and employs a set of 92 bacterial core genes that are single-copy and commonly present in all bacterial genomes.

The main characteristics of the genomes are presented in Table 1. The draft genome of strain CECT 9293^T has an estimated size of 4.35 Mbp. It is composed of 149 contigs with an N50 value of 84 805 nucleotides and a final assembly coverage of 283×. CheckM results of contamination and completeness were 0.98 and 99.9 %, respectively. The assembly contains 3598 protein-coding sequences and 53 RNA genes. The G+C content is 32.6 mol%. Only one rRNA operon is detected and its 16S rRNA gene sequences is complete and 100 % coincident with the partial sequence previously amplified by Sanger technology. The genome of strain CECT 9390^T has an estimated size of 3.49 Mbp. It is composed of two contigs with an N50 value of 3462687 nucleotides and final assembly coverage of 2087×. It should be noted that strain CECT 9390^T was sequenced through PacBio technology, thus its genome is closed and could be resolved as containing one chromosome of 3.46 Mb and a 27.8 Kb independent contig, assumed to be a plasmid. CheckM results of contamination and completeness were 0.49 and 100 %, respectively. The assembly contains 3203 protein-coding sequences and 67 RNA genes. Four rRNA operons are detected and its 16S rRNA gene sequence is complete and 100 % coincident with the partial sequence previously amplified by Sanger technology. The G+C molar content is 34.5 mol%. Finally, the draft genome of strain CECT 9393^T has an estimated size of 4.04 Mbp. It is composed of 149 contigs with an N50 value of 35 975 nucleotides and final assembly coverage of 126×. CheckM results of contamination and completeness were 0.49 and 100 %, respectively. The assembly contains 3433 protein-coding sequences and 53 RNA genes. The G+C molar content is 32.7 mol%. Only one rRNA operon is detected and its 16S rRNA gene sequences is complete and 100 % coincident with the partial sequence previously amplified by Sanger technology.

Table 1.

Characteristics of the genomes of the drinking water Chryseobacterium strains analysed in this study (in bold) with their close type strains and some other reference organisms

Strain	Size (Mb)	G+C (mol%)	Coverage (×)	N50 (Kpb)	Contigs	Protein	rRNA	tRNA	Accession numbers*
C. aquaeductus CECT 9390T	3.49	34.5	2087	3 462 687	2†	3203	12	55	GCF_905175375 (CAJIMS01)
C. potabilaquae CECT 9293T	4.35	32.6	283	84 805	149	3598	3	50	GCA_902728265 (CACVBR01)
C. fistulae CECT 9393T	4.04	32.7	126	35 975	233	3433	3	50	GCA_902729325 (CACVBY01)
C. aquaticum KCTC 12483T	3.81	33.9	125	465 295	21	3387	6	64	GCA_001420285 (LLYZ01)
C. artocarpi UTM3T	4.94	34.8	184	718 111	51	4432	8	79	GCF_001684975 (MAYH01)
C. daecheongense DSM 15235T	3.83	36.2	274	892 685	9	3490	3	60	GCA_004365465 (SOQW01)
C. defluvii DSM 14219T	3.71	36.6	284	1 111 720	8	3421	3	65	GCA_003634775 (RBXB01)
C. gleum ATCC 35910T	5.57	36.8	31.57	3 504 888	7	4994	5	41	GCA_000143785 (ACKQ02)
C. oncorhynchi 701B08T	4.79	35.1	82.82	735 093	47	4331	8	76	GCF_002899895 (PPEI02)
C. piscicola DSM 21068T	3.45	33.8	364	311 216	28	3000	3	49	GCA_900156685 (FTOJ01)
C. viscerum 687B08T	5.69	36.2	61.09	543 073	60	4997	3	73	GCF_002899945 (PPEG02)
E. hispanica KCTC 22104T	3.78	34.6	35.9	89 519	149	3434	10	46	GCF_003385395 (QNUG01)

*Assembly accession shown first with WGS project in parentheses.

†Chromosome (3.46 Mb) and plasmid (27.8 Kb).

Table 2 shows the values obtained for the three overall genomic relatedness indexes that were explored, namely ANIb, isDDH and AAI, relating the genomes of the three new strains with those of the type strains of neighbouring, validly named species and references used for comparison. The first noticeable finding is that the three strains show ANIb and isDDH values qualifying each of them as different species. The overall phenotypic resemblance and the high 16S rRNA gene sequence similarity between strains CECT 9293^T and CECT 9393^T combined with an ANIb value of 92 % and 51.5 % isDDH, indicate they clearly pertain to close, but different, yet unnamed species. All other values on Table 2 are lower than 86 % for ANIb and 31 % for isDDH. AAI values follow the same tendency, with maximum figures of 94 % for the pair CECT 9293^T and CECT 9393^T, followed by a 91 % between C. daecheongense and C. defluvii type strain genomes. It is noticeable that all intrageneric AAI values relating Chyseobacterium species are ≥74 %, while values among them and species of the neighbouring genera Epilithonimonas , Halpernia and Kaistella lay below 70 % meaning that they are well demarcated.

Table 2.

Overall genome relatedness indexes between strains CECT 9390^T, CECT 9293^T and CECT 9393^T (in bold) to nearest Chryseobacterium species and to the type species of the nearest genus Epilithonimonas

(a) ANIb* and isDDH (blue); (b) AAI (green, intergeneric values).

(a)	1	2	3	4	5	6	7	8	9	10	11	12
1. C. aquaeductus CECT 9390T		20.3	20.3	26.1	21.5	20.9	20.8	21,3	21.2	23.5	21.6	22.0
2. C. potabilaquae CECT 9293T	75.9		51.5	20.9	21.7	21.5	21.1	21,7	21.5	20.5	21.4	21.0
3. C. fistulae CECT 9393T	76.0	92.1		21.6	21.8	21.8	21.3	21,9	21.6	21.0	21.7	27.1
4. C . aquaticum KCTC 12483T	82.3	76.3	76.7		21.4	21.4	21.4	21,7	21.6	22.9	22.9	23.7
5. C . artocarpi UTM3T	76.7	77.0	77.0	76.8		22.6	22.0	25,3	27.0	21.0	25.5	20.9
6. C . daecheongense DSM 15235T	76.8	77.4	77.5	77.2	78.1		30.8	23,2	22.4	20.4	22.3	20.8
7. C . defluvii DSM 14219T	76.7	77.2	77.2	77.1	78.0	85.7		22,8	22.1	20.8	22.3	20.5
8. C. gleum ATCC 35910T	76.2	76.5	76.9	76.3	80.8	78.4	78.1		25,3	21,1	30,0	22,7
9. C . oncorhynchi 701B08T	76.5	77.0	77.1	76.8	82.9	78.1	78.1	81.2		20.8	26.3	21.0
10. C . piscicola DSM 21068T	78.6	75.1	75.3	78.5	75.5	75.5	75.6	75.2	75.5		21.4	21.7
11. C . viscerum 687B08T	76.4	76.3	76.6	77.3	81.1	77.9	77.9	84.2	82.0	75.5		23.3
12. E. hispanica KCTC 22104T	73.4	72.2	73.2	74.1	72.3	72.4	72.7	72.9	72.4	73.3	72.7
(b)	1	2	3	4	5	6	7	8	9	10	11	12
1. C. aquaeductus CECT 9390T
2. C . aquaticum KCTC 12483T	85
3. C . piscicola DSM 21068T	79	79
4. C. potabilaquae CECT 9293T	76	77	75
5. C. fistulae CECT 9393T	77	77	75	94
6. C . defluvii DSM 14219T	78	78	76	80	80
7. C . daecheongense DSM 15235T	78	78	76	80	80	91
8. C . oncorhynchi 701B08T	77	76	75	78	79	80	80
9. C . artocarpi UTM3T	76	76	74	78	79	80	80	87
10. C . viscerum 687B08T	76	77	74	78	78	79	79	85	84
11. C . gleum ATCC 35910T	76	77	75	79	80	80	81	85	84	88
12. E. hispanica KCTC 22104T	68	69	68	68	69	68	68	67	67	67	69

*ANIb figures are median values among reciprocal ANIb.

†Intergeneric AAI values between Chryseobacterium species, four Epilithonimonas species and three Kaistella species are always ≤70 % (not shown).

Phylogenomic relationships were explored using UBCG-based trees, which are based on 92 single-copy, universally distributed bacterial genes. Trees generated by using nucleotide and aminoacidic sequences were generated, including 63 genomes of Chryseobacterium type strains plus genomes of Epilithonimonas, Halpernia, Kaistella , Weeksella and Flavobacterium type strains. The results are depicted in Fig. 2 (aminoacid-based UBCG tree) and Fig. S1 (available in the online version of this article; nucleotide-based UBCG tree). Both trees show the position of strain CECT 9390^T paired with the type strains of C. aquaticus and C. piscicola , while strains CECT 9293^T and CECT 9393^T, paired between them, are closest to C. daecheongense and C. defluvii , all of them in the core of the genus Chryseobacterium . The genus boundaries are well defined from genera recently detached from it, coincident with the findings of Nicholson et al. [5]. It is interesting to highlight that relationships seen on the 16S rRNA gene-based tree do not coincide with the ones revealed through the phylogenomic UBCG tree: for example, the closest neighbour of CECT 9390^T is C. glaciei with 16S rRNA data (highest similarity, closest neighbour), but C. aquaticum from genome data. Aside from the important difference in information content between both datasets, quality issues with 16S rRNA sequences might also account for the results. For example, the 16S rRNA of C. glaciei IHBB 10212^T (KR233779) was not entirely identical to the sequences in its genome assembly (GCF_001648155), sharing only 99.7 % with its seven copies.

Fig. 2.

Phylogenetic tree generated with the UBCG [22] by using amino acid sequences. The numbers at the nodes indicate the gene support index (maximal value is 92). Genome accession numbers are indicated in parentheses. Bar, 0.05 substitutions per position.

Ecology

A search for potential additional strains or uncultivated representatives of these three new species was performed with blast using complete 16S rRNA gene sequences derived from genomes: CECT 9390^T sequence had one hit with 99.3 % similarity to strain E-052908, isolated from surface of an historic monument in Scotland [23], all other were less than 99 %. The EzBioCloud search rendered no hit over 99 %. The CECT 9293^T and CECT 9393^T sequences had no hits higher than 99.0 % (except to each other) when submitted to blast, but when the EzBiocloud search was applied, they gave hits of 99.6–99.8 % to the genome sequence of strain ‘Chryseobacterium nematophagum’ Jub275, which is able to kill the nematode Caenorhabditis elegans [24]. Further comparison of strains CECT 9293^T and CECT 9393^T with the two strains studied by Page et al. [24], i.e. Jub129 and Jub275 (GCA_003710065.1 and GCA_003709475.1), was performed at the genomic level. The highest isDDH value found related strains CECT 9293^T and Jub275 by 63.4 %, followed by 60.8 % between CECT 9393^T and Jub129. The corresponding ANIb values were 94.8 and 93.7 %, respectively. In all cases, the figures were lower than the thresholds for species differentiation (70 % for DDH and 95–96 % for ANI). Strains Jub129 and Jub275 were isolated from nematodes (Caenorhabditis briggsae) found on rotten fruits (apple, fig), an environment not resembling at all the one of strains CECT 9293^T and CECT 9393^T. Interestingly, despite the assumption of Page et al. that strains Jub129 and Jub275 belong to the same species, our ANI and isDDH data for these two strains reveal that they represent two different genomic species (55.5 % isDDH and 93.4 % ANIb). In summary, we have not found strains or uncultured clones putatively corresponding to the same species in 16S rRNA gene databases and we are thus unable to define the distribution of the species that seem to be rare and not abundant.

Phenotypic characterization

In order to perform a comparative analysis of the phenotypic profile of the drinking water strains, Chryseobacterium aquaticum CECT 7302^T and C. piscicola CECT 7357^T were characterized in parallel. Four more strains, C. artocarpi CECT 8497^T, C. oncorhynchi CECT 7794^T, C. viscerum CECT 7793^T and Epilithonimonas hispanica CECT 7129^T, were also tested as control but their results are not included or discussed as they are not close relatives of any of the novel species. Procedures for testing phenotyping traits followed various published recommendations [1, 2, 25]. Unless otherwise indicated, the strains were cultured on Reasoner's 2A (R2A) medium (Difco) at 26 °C.

Cell morphology was determined on wet mounts prepared from 48 h cultures on R2A agar, by using phase contrast microscopy in a Leica DMRB fluorescence microscope (Fig. S2). Colony morphology and pigmentation were recorded from 48 h R2A and TSA (tryptone soy agar; Difco) plate cultures. Flexirubin-type pigment production was tested as described by Bernardet et al. [1]. Temperature (4, 15, 26, 30, 37 and 40 °C) and pH (pH 4–10.5, at 0.5 pH unit intervals) ranges for growth were determined on R2A broth (0.5 g yeast extract, 0.5 g protease peptone, 0.5 g casein hydrolysate, 0.5 g glucose, 0.5 g soluble starch, 0.3 g sodium pyruvate, 0.3 g K₂HPO₄, 0.05 g Mg SO₄·7H₂O, 12.0 g agar powder) incubated for 48 h (up to 7 days at 15 and 4 °C). Optimal values were taken from the fastest grown tubes. Ability to grow in tryptone soy broth (TSB; Difco) plus 3 % NaCl and salinity range for growth (0–1.0, 1.5–2.0, 2.5–3.0, 4.0–5.0 and 6.0 % sea salts in 5 g l^–1 tryptone and 1 g l^–1 yeast extract) were determined after 3 days incubation. Growth on cetrimide agar (Pronadisa), marine agar 2216 (MA; Difco) and MacConkey agar (Pronadisa) were tested after 2 days incubation at 26 °C. Extracellular hydrolytic activities on casein and starch were determined on TSA supplemented with 10 % (v/v) casein suspension or 0.2 % (w/v) soluble starch, respectively. Tween 80 agar [26] and DNAse agar (Oxoid) were used to detect esterase and DNase activities. Activity on starch was revealed after lugol addition, and 1 M HCl was used to show DNase activity. Oxidase test was performed with Oxoid oxidase discs and catalase was tested with 10 vol% H₂O₂. API 20E, API 20NE, API 50CHE and API ZYM strips were inoculated following the manufacturer’s instructions. Oxidation/fermentation medium of Hugh and Leifson (Difco) was used to further test the aerobic acidification of the following carbohydrates: d-glucose, d-fructose, trehalose, d-xylose, maltose, cellobiose, sucrose, d-glycerol and d-mannitol, by incubating the tubes at 26 °C for 6 days.

Fatty acid methyl esters were extracted from biomass grown for 48 h on WiPCA at 26 °C and prepared according to standard protocols as described for the midi Microbial Identification System [27]. Cellular fatty acid content was analysed by gas chromatography with an Agilent 6850 chromatographic unit, with the midi Microbial Identification System using the TSBA6 method [28] and identified using the Microbial Identification Sherlock software package.

Results of the phenotypic analysis are reported in the species descriptions and in Table 3, which displays the differential features between each strain and their phylogenetically closest neighbours. All three strains fit the defining characteristics of the genus Chryseobacterium according to the latest emendation by Nicholson et al. [5]. In some instances, traits not determined experimentally, such as major polar lipids, could be inferred from gene content; thus, the presence of a phosphatidyl serine decarboxylase gene points to the ability to synthesize phosphatidyl ethanolamine, the major polar lipid in most Chryseobacterium [29]. The presence of hexaprenil diphosphate synthase genes (but not others for larger isoprene chain) as part of the subsystem for ‘isoprenoid for quinones’ suggests MK-6 as major quinone. The search for homospermidine synthase gene, whose product is responsible for sym-homospermidine synthesis (the main polyamine of the genus according to [30]) gave no result. It is also noticeable the presence of several genes involved in capsular and extracellular polysaccharide synthesis in all three genomes, in agreement with the very mucous colony types that the strains display. Genes for β-lactamases, chloramphenicol acetyltransferase and RND efflux systems are present in all three strains, while genes for MATE family MDR pump occur only in strains CECT 9293^T and CECT 9393^T.

Table 3.

Differential characteristics between strains CECT 9390^T, CECT 9293^T, CECT 9393^T and their closest phylogenomic relatives

Strains: 1, Chryseobacterium aquaeductus CECT 9390^T; 2, Chryseobacterium aquaticum CECT 7302^T; 3, Chryseobacterium piscicola CECT 7357^T; 4, Chryseobacterium potabilaquae CECT 9293^T; 5, Chryseobacterium fistulae CECT 9393^T; 6, Chryseobacterium defluvii B2^T [2]; 7, Chryseobacterium daecheongense CPW 406^T [2]. +, Positive; −, negative; w, weak; nd, not determined. All data from this study unless otherwise indicated.

Characteristic	1	2	3	4	5	6	7
Growth on/at:
TSB 3 % NaCl	+	+	−	−	−	−	−
MA	+	+	−	−	−	nd	nd
Cetrimide agar	−	−	−	−	−	+	+
5 °C	+	−	−	−	−	−	−
37 °C	−	+	−	−	−	+	+
Nitrate reduction to nitrite	−	−	−	−	+	−	+
DNAse	−	+	−	−	−	+	+
Urease	−	−	−	+	+	−	−
N-Acetyl-β-glucosidase	−	+	w	−	−	nd	nd
Acid from:
d-Glucose	−	−	−	+	+	+	−
d-Fructose	−	−	−	−	−	−	+
Trehalose	−	+	−	+	+	+	+
d-Xylose	−	−	−	−	−	−	+
Maltose	−	−	−	−	−	+	−
Cellobiose	−	−	−	−	−	+	+
Sucrose	−	+	−	−	−	−	−
d-Glycerol	−	−	−	−	−	−	+
Genomic traits:
β-Carotene 3-hydroxylase gene	+	+	+	−	+	+	+
CRISPR-Cas systems	−	+ (Cas 1,2,9)	+ (Cas 1,2,9)	+ (Cas 1,2,9)	−	+ (Cas 1,2)	−
Restriction−modification systems	−	+	+	−	+	+	+

Regarding the fatty acid composition (Table S1), all three strains had C_15 : 0 iso as a main component followed by C_17 : 0 iso 3-OH, but in a rather wide range of abundance (24.0–54.8 % and 10.8–17.5 %, respectively). Other fatty acids added yet more differences between the strains both in presence and relative amount.

Strain 9390^T could be easily differentiated from their closest relatives, C. aquaticum and C. piscicola , by a combination of temperature growth range, DNAse and N-acetyl-β-glucosidase activities, ability to produce acid aerobically from trehalose and sucrose, and to grow in TSB 3 % NaCl and MA. Differentiation between 9293^T and 9393^T relys only in one characteristic, among the many tested: nitrate reduction ability, but they are differentiated from both C. defluvii and C. daecheongense by at least five traits; another additional six features allow further differentiation from each of the two relatives, as shown in Table 3. Genomic information adds some other distinctive characters, as the presence of some of the genes involved in carotenoid synthesis or the presence of restriction–modification systems and CRISPR-Cas proteins.

Taken together, the results under this heading provide phenotypic support for the separate species status that was shown by the phylogenetic analysis. Therefore, the assignment of each of the three water system isolates, strains CECT 9293^T, CECT 9390^T and CECT 9393^T, to novel species is proposed with the following descriptions.

Description of Chryseobacterium potabilaquae sp. nov.

Chryseobacterium potabilaquae (po.ta.bil.a'quae. L. masc. adj. potabilis potable; L. fem. n. aqua water; N.L. gen. n. potabilaquae of drinking water).

Cells are bacilli with slightly pointed ends, 1.5–3.5 µm×0.5–0.6 µm in size, Gram-reaction-negative and non-motile. Colonies on R2A medium are elevated, mucous and pale yellow. Grows on TSA, but growth is better and faster on R2A medium. Flexirubin-type pigments are produced, according to the KOH test. Does not grow on MacConkey agar, cetrimide agar, MA or in TSB plus 3 % (w/v) NaCl. Salinity tolerance range for growth is 0–2.5 %. Oxidase and catalase tests are positive. Temperature range for growth is 15–30 °C, no growth is observed at 4 or 37 °C. The pH range for growth is pH 5.0–8.5, with slight growth at pH 9.0. Aerobic, unable to ferment carbohydrates, but acid is produced (weakly) in aerobic conditions from d-glucose and trehalose. Nitrate is not reduced to nitrite or further. The following activities are negative: arginine dihydrolase, lysine decarboxylase, ornithine decarboxylase, β-galactosidase, tryptophanase (indole), phenylalanine deaminase, Voges–Proskauer and acid production from d-glucose, d-mannitol, d-sorbitol, myo-inositol, l-rhamnose, sucrose, melibiose, amygdalin and l-arabinose on API 20E; positive for urea, aesculin and gelatin hydrolysis, alkaline phosphatase, leucine and valine arylamidases, trypsin and naphthol-AS-BI-phosphohydrolase; weak positive response for acidic phosphatase. Hydrolyses casein, Tween 80 and starch (weak), but not DNA. Tyrosine clearing is negative but growth on tyrosine medium produces a diffusible brown pigmentation. No acidification is observed in API 50CH/E, except for a slight response on d-glucose, maltose and trehalose. Acid production in aerobic oxidation/fermentation test is positive for d-glucose and trehalose but negative for d-xylose, l-arabinose, d-fructose, maltose, cellobiose, sucrose, d-glycerol and d-mannitol. Major fatty acids are C_15 : 0 iso and C_17 : 0 iso 3-OH, followed by C_16 : 1  ω7c/C_16 : 1 ω6c, C_16 : 0 10-methyl/C_17 : 1 iso ω9c and C_15 : 0 anteiso.

The type strain is CECT 9293^T (=LMG 32084^T), which was isolated from a drinking water treatment plant in Barcelona, Spain.

Genome of the type strain is 4.35 Mb in size and comprises 3598 protein-coding genes and 53 RNA genes (a single rRNA operon and 50 tRNAs). The G+C content is 32.6 mol%.

Description of Chryseobacterium aquaeductus sp. nov.

Chryseobacterium aquaeductus (a.quae.duc’tus. L. gen. n. aquaeductus of a conveyance of water).

Cells are Gram-reaction-negative, short, non-motile bacilli, 0.8–2.0 µm×0.6–0.8 µm, that do not glide. Colonies on R2A medium are yellow, mucoid and shiny. The pigmentation on TSA medium is ochre. Flexirubin-type pigments are produced, according with the KOH test. Chemoorganotrophic, strictly aerobic, positive for catalase and oxidase tests. Carbohydrates are not fermented. Grows on MA after 4 days but not on MacConkey or cetrimide agar. Grows in TSB with 3 % NaCl. Salinity range of growth is 0–3.0 %. Temperature range for growth is 4–30 °C, no growth is observed at 37 °C or more. pH range for growth is pH 5.5–9.5. Negative for nitrate reduction to nitrite (or further), arginine dihydrolase, lysine decarboxylase, ornithine decarboxylase, urease, β-galactosidase, tryptophanase (indole), phenylalanine deaminase, Voges–Proskauer and acid production from d-glucose, d-mannitol, d-sorbitol, myo-inositol, l-rhamnose, sucrose, melibiose, amygdalin and l-arabinose; positive for aesculin and gelatin hydrolysis, alkaline phosphatase, leucine and valine arylamidases and trypsin; weak positive response for esterase lipase, acidic phosphatase and naphthol-AS-BI-phosphohydrolase. Hydrolyses casein, Tween-80 and starch, but not DNA. Tyrosine clearing is negative but growth on tyrosine medium produces a strong diffusible brown pigmentation. No acidification is observed in any of the 49 carbohydrates of API 50CH/E, with only a very slight response in maltose, trehalose, amygdalin and d-gentibiose. No acidification is observed in aerobic oxidation/fermentation medium with any of the following carbohydrates: d-xylose, l-arabinose, d-glucose, d-fructose, trehalose, maltose, cellobiose, sucrose, d-glycerol and d-mannitol. Major fatty acids are C_15 : 0 iso, C_17 : 0 iso 3-OH and C_16 : 1  ω7c/C_16 : 1 ω6c, followed by C_15 : 1 ω5c and C_15 : 0 anteiso.

The type strain is CECT 9390^T (=LMG 32085^T), which was isolated from a drinking water distribution network in Barcelona, Spain.

Genome of the type strain is 3.49 Mb in size (one chromosome of 3.46 Mb and one plasmid of 27.8 Kb) and comprises 3203 protein-coding genes plus 68 RNA genes. It contains four rRNA operons and 55 tRNA genes. The G+C content is 34.5 mol%.

Description of Chryseobacterium fistulae sp. nov.

Chryseobacterium fistulae (fis'tu.lae. L. gen. n. fistulae of a water pipe).

Cells are bacilli with slightly pointed ends, 1.5–3.5 µm×0.5–0.6 µm in size, Gram-reaction-negative and non-motile. Colonies on R2A medium are elevated, mucous and pale yellow. Grows on TSA, but growth is better and faster on R2A medium. Flexirubin type pigments are produced, according with the KOH test. Does not grow on MacConkey agar, cetrimide agar or MA or in TSB plus 3 % (w/v) NaCl. Salinity tolerance range for growth is 0–2.0 %. Temperature range for growth is 15–30 °C, no growth is observed at 4 or 37 °C. The pH range for growth is pH 5.0–8.5, with slight growth at pH 9.0. Oxidase and catalase tests are positive. Aerobic, unable to ferment carbohydrates, but acid is produced (weakly) in aerobic conditions from d-glucose and trehalose. Nitrate is reduced to nitrite. The following activities are negative: arginine dihydrolase, lysine decarboxylase, ornithine decarboxylase, β-galactosidase, tryptophanase (indole), phenylalanine deaminase, Voges–Proskauer and acid production from d-glucose, d-mannitol, d-sorbitol, myo-inositol, l-rhamnose, sucrose, melibiose, amygdalin and l-arabinose on API 20E; positive for urea, aesculin and gelatin hydrolysis, alkaline and acidic phosphatases, leucine and valine arylamidases, trypsin, and naphthol-AS-BI-phosphohydrolase. Hydrolyses casein, Tween-80 and starch (weak), but not DNA. Tyrosine clearing is negative but growth on tyrosine medium produces a diffusible brown pigmentation. No acidification is observed in API 50CH/E, except for a slight response on trehalose. Acid production in aerobic oxidation/fermentation (O/F) test is positive for d-glucose and trehalose but negative for d-xylose, l-arabinose, d-fructose, maltose, cellobiose, sucrose, d-glycerol and d-mannitol. Major fatty acids are C_15 : 0 iso, C_17 : 0 iso 3-OH, C_16 : 0 10-methyl/C_17 : 1 iso ω9c and C_15 : 0 anteiso followed by C_16 : 1  ω7c/C_16 : 1 ω6c and C_15 : 1 ω5c.

The type strain is CECT 9393^T (=LMG 32086^T), which was isolated from a drinking water distribution network at Barcelona, Spain.

The genome of the type strain is 4.04 Mb in size and comprises 3433 protein-coding genes and 53 RNA genes (one rRNA operon and 50 tRNAs). The G+C content is 32.7 mol%.