Description of Chryseobacterium anthropi sp. nov. to accommodate clinical isolates biochemically similar to Kaistella koreensis and Chryseobacterium haifense, proposal to reclassify Kaistella koreensis as Chryseobacterium koreense comb. nov. and emended description of the genus Chryseobacterium

Within the family Flavobacteriaceae as emended by Bernardet et al. (2002), the genera Bergeyella, Chryseobacterium, Elizabethkingia, Empedobacter, Kaistella, Riemerella, Sejongia and Weeksella, and the recently described genus Wautersiella (Kämpfer et al., 2006) form a distinct branch. The genus Chryseobacterium currently contains more than 20 species, including some of clinical importance (Quan et al., 2007; Vaneechoutte et al., 2007). The genus Kaistella was proposed by Kim et al. (2004) as a novel member of the Chryseobacterium–Bergeyella–Riemerella branch, mainly on the basis of 16S rRNA gene sequence comparisons, without reporting striking phenotypic features (i.e. chemotaxonomic traits) as differentiating characteristics. Interestingly, the species Chryseobacterium haifense (Hantsis-Zacharov & Halpern, 2007) clearly shows higher 16S rRNA gene sequence similarity to the sole species of Kaistella (97 %) than to the Chryseobacterium species described at that time. However, these authors did not compare C. haifense to Kaistella koreensis in their paper.

Here, we report the characterization of eight strains of clinical origin (Table 1⇓) that were biochemically most similar to K. koreensis and C. haifense and which displayed biochemical profiles that resembled those of CDC group II-e strains (Dees et al., 1986).

Biochemical and morphological tests were performed as described previously (Laffineur et al., 2002; Schreckenberger et al., 2003; Vaneechoutte et al., 2007) on the eight clinical isolates and on the other strains listed in Table 2⇓. Assimilation and alkalinization of organic compounds was detected on Simmons' citrate agar base, replacing citrate with various organic substrates at 0.2 % (w/v), according to Martin et al. (1981). Enzyme activities were detected using diagnostic tablets (Rosco). The KOH test was used to detect flexirubin pigments (Bernardet et al., 2002). Table 2⇓ summarizes the biochemical data obtained. All eight strains grew aerobically at 37 °C (with optimal growth at 30 °C) and were positive for oxidase and catalase activities and for acid production from glucose and maltose. Acid production from mannose and hydrolysis of Tween 80 were strain-dependent. All strains were positive for indole production, hydrolysis of gelatin, starch and tyrosine, presence of alkaline phosphatase, trypsin and pyrrolidonyl aminopeptidase activities and alkalinization of acetate. All strains were negative for production of flexirubin pigments, growth at 42 °C, nitrate reduction, H₂S production on Kligler agar, alkalinization of citrate and urease, β-galactosidase, ornithine and lysine decarboxylase, arginine dihydrolase and l-phenylalanine deaminase activities. The phenotypic characteristics most useful to differentiate the eight clinical isolates from related taxa are highlighted in Table 2⇓.

In summary, the strains can be differentiated from K. koreensis (Kim et al., 2004) by alkalinization of acetate, from C. haifense (Hantsis-Zacharov & Halpern, 2007) by the lack of acid production from fructose, lactose and sucrose and by a negative ONPG test (β-galactosidase), from Chryseobacterium molle and Chryseobacterium pallidum (Herzog et al., 2008) by the lack of acid production from arabinose, cellobiose, xylose and ethylene glycol and by the lack of hydrolysis of aesculin, from Chryseobacterium hispanicum (Gallego et al., 2006) and Chryseobacterium caeni (Quan et al., 2007) by the absence of flexirubin pigments and by the lack of aesculin hydrolysis and from Chryseobacterium hominis (Vaneechoutte et al., 2007) by the lack of aesculin hydrolysis and the lack of acid production from ethylene glycol.

Antibiotic susceptibility was determined by the disc diffusion method using the criteria for non-Enterobacteriaceae Gram-negative rods (CLSI, 2005). All strains were susceptible to erythromycin, ciprofloxacin, temocillin and trimethoprim-sulfamethoxazole. All but one were susceptible to tetracycline. Susceptibility to desferrioxamine, ampicillin, cephalothin, cefotaxime and gentamicin was strain-dependent.

In contrast to the original description of K. koreensis (Kim et al., 2004), where oxidation–fermentation of glucose, using the medium of Hugh & Leifson (1953), tested negative, we found K. koreensis KCTC 12107^T to be saccharolytic, producing acid from glucose, maltose and mannose, by three different methods, using (i) ammonium salt agar base, which is the method recommended by Bernardet et al. (2002), (ii) the conventional oxidation–fermentation medium of Hugh & Leifson (1953) and (iii) low-peptone phenol red agar base (Wauters et al., 1998). Furthermore, we could not confirm the hydrolysis of aesculin.

Our results for C. caeni N4^T are also significantly different from the original description (Quan et al., 2007). The authors described an asaccharolytic species, whereas we found the type and only strain of this species to be one of the most saccharolytic strains among the Chryseobacterium species, rapidly acidifying glucose, maltose, sucrose, trehalose and l-arabinose and acidifying xylose and cellobiose more slowly. This contradiction may be explained by the fact that the authors used three API strips for carbohydrate metabolism, API 20 E, API 20 NE and API ID 32 GN (bioMérieux), although these are not suitable for the detection of oxidative acidification of carbohydrates by non-fermenters: API 20 E is intended primarily for fermentative bacteria and is not reliable for non-fermenters, and API 20 NE and API ID 32 GN only score assimilation, which is often different from acidification. Moreover, indole production was reported as negative, whereas the strain was clearly positive in our hands.

16S rRNA gene sequences were determined for all eight isolates, as described previously (Wauters et al., 2003), and a phylogenetic tree was constructed on the basis of the 16S rRNA gene sequences using the software package Genebase (Applied Maths). At first, a pairwise alignment was carried out using UPGMA (gap cost 100 %, unit gap cost 20 %), which was subsequently used for a global alignment with Bacteroides fragilis ATCC 25285^T as an outgroup on a common region corresponding to Escherichia coli 16S rRNA positions 211–1378, resulting in a similarity matrix. A dendrogram was constructed on the basis of this similarity matrix, using the neighbour-joining clustering method (Fig. 1⇓). Bootstrap values, obtained from 1000 replicated calculations, are presented as percentages at each node, omitting values below 50 %. The similarity between strain NF 1366^T and the seven other strains ranged between 99.7 and 100 %. Based on 16S rRNA gene sequence similarity, this collection of strains was most closely related to C. haifense DSM 19056^T (99.1 %) and K. koreensis KCTC 12107^T (97.3–97.5 %). One sequence deposited in GenBank (AB035150) showed more than 99.7 % similarity to the sequence of NF 1366^T and was listed as ‘Haloanella gallinarum’, a name not encountered among validly published species names. Apparently, this sequence was obtained from a strain that belongs to the same species as NF 1366^T.

The G+C content of the DNA of strain NF 1366^T was determined at the DSMZ (Braunschweig, Germany), according to the method of Mesbah et al. (1989), as being 39.0 mol%. This is close to the values of 37.8 mol% reported for C. haifense DSM 19056^T (Hantsis-Zacharov & Halpern, 2007) and 41.2–41.6 mol% for K. koreensis KCTC 12107^T (Kim et al., 2004).

tRNA-intergenic length polymorphism analysis (tDNA-PCR) was carried out as described previously (Baele et al., 2000). The tDNA-PCR patterns of the eight strains studied were compared with patterns obtained from all strains of the family Flavobacteriaceae available in the collection of the LBR UGent. A similarity tree (Fig. 2⇓) was constructed as follows: the distance matrix was calculated with the differential base pairs algorithm, using a 1 bp tolerance threshold (Baele et al., 2001), clustering was done with the UPGMA algorithm using the neighbor software and the tree was constructed with the phylip software (TreeView; ). Two of the strains studied, NF 450 and CCUG 15624, showed a pattern that was intermediate between that of the six other strains and that of C. haifense DSM 19056^T. All eight strains studied and C. haifense DSM 19056^T had a tRNA intergenic spacer with a length of 135.2±0.1 bp (mean±sd) in common, all eight strains studied had fragments of 106.9±0.1 and 205.9±0.1 bp in common and the two intermediate strains had spacers of 210.4, 229.1 and 243.1 bp in common with C. haifense DSM 19056^T, but could be differentiated from it by the presence of fragments of 108.3, 109.1 and 205.1 bp in C. haifense DSM 19056^T. The two intermediate strains could be differentiated from the other six strains studied by the presence of fragments of 212.0, 230.3 and 244.6 bp in the latter. K. koreensis KCTC 12107^T had tRNA spacer fragments of 101.0, 144.5, 197.6, 211.8, 245.3 and 284.2 bp.

For DNA–DNA hybridization, a microtitre plate method was used, modified after Lind & Ursing (1986), as described previously (Ziemke et al., 1998). Strain NF 1366^T showed only 41.8 % DNA–DNA relatedness with K. koreensis DSM 12107^T and only 51.9 % with C. haifense DSM 19056^T, and therefore was considered to represent a separate species. Strains NF 450 and CCUG 15624 were phenotypically identical to the six other strains, but their tDNA-PCR pattern was intermediate between that of C. haifense DSM 19056^T and those of the six other strains. The DNA–DNA relatedness of these two strains was 41.7 and 74.6 % with strain NF 1366^T and 72.6 and 70.2 % with C. haifense DSM 19056^T, respectively. Hence, both strains must be classified as intermediate between C. haifense and the novel species described below, in agreement with their intermediate tDNA-PCR pattern.

Biomass used for the extraction of quinones, polar lipids and polyamines was obtained after culture at 28 °C in 3.3× PYE medium (10 g peptone from casein and 10 g yeast extract l⁻¹, pH 7.2). Polar lipids were extracted and analysed as described previously (Tindall, 1990; Altenburger et al., 1996). Polyamines were extracted and analysed as reported by Busse & Auling (1988) and Stolz et al. (2007). The polyamine pattern consisted of the major compound sym-homospermidine [36.5 μmol (g dry weight)⁻¹], minor amounts of spermidine [2.9 μmol (g dry weight)⁻¹] and spermine [3.1 μmol (g dry weight)⁻¹] and traces of putrescine [<0.1 μmol (g dry weight)⁻¹], which is consistent with the characteristics of other chryseobacteria (Hamana & Matsuzaki, 1990, 1991; Kämpfer et al., 2003). It is worth mentioning that, in the original genus description, major amounts of 2-hydroxyputrescine were also reported in Chryseobacterium indologenes (Vandamme et al., 1994; Hamana & Matsuzaki, 1991). However, the diamine 2-hydroxyputrescine has so far been reported only from betaproteobacteria whereas, for all chryseobacteria for which polyamine patterns have been determined, including the type species of the genus, Chryseobacterium gleum, and the species Chryseobacterium balustinum, C. daecheongense, C. defluvii, C. hispanicum, C. indoltheticum and C. scophthalmum, only sym-homospermidine has been reported as the major compound (Hamana & Matsuzaki, 1990, 1991; Kämpfer et al., 2003; Kim et al., 2005; Hamana et al., 2008). Possibly, 2-hydroxyputrescine was identified erroneously in C. indologenes. Hence, the genus description of Chryseobacterium should be emended to indicate that members of the genus contain sym-homospermidine as the major polyamine.

The quinone system of strain NF 1366^T consisted exclusively of menaquinone MK-6, which also is in agreement with the assignment of NF 1366^T to the genus Chryseobacterium (Vandamme et al., 1994). The polar lipid profile of NF 1366^T exhibited the major compounds phosphatidylethanolamine, an unknown aminophospholipid, an unknown aminolipid and two unknown polar lipids and minor to trace amounts of four unknown aminolipids and seven unknown polar lipids (Fig. 3a⇓). This polar lipid profile is almost indistinguishable from that of C. gleum (data not shown) and is also similar to that of C. defluvii, except that no rather hydrophilic aminophospholipid was detected in NF 1366^T.

The polar lipid profile of K. koreensis KCTC 12107^T consisted of the predominant compounds phosphatidylethanolamine, an unknown aminolipid (AL1) and three unknown polar lipids L2, L3 and L4, with the first two exhibiting a very weak reaction with ninhydrin, indicating that they may contain amino groups (Fig. 3b⇑). Furthermore, moderate amounts of an unknown polar lipid L6 and an aminophospholipid APL1 and minor amounts of three unknown polar lipids and three aminolipids could be detected. So far, polar lipid analysis has been carried out for the species C. defluvii (Kämpfer et al., 2003), C. daecheongense (Kim et al., 2005), C. aquaticum (Kim et al., 2008) and C. gambrini, C. molle, C. pallidum and C. ureilyticum (Herzog et al., 2008). Unfortunately, only the papers describing C. defluvii, C. gambrini, C. molle, C. pallidum and C. ureilyticum (Kämpfer et al., 2003; Herzog et al., 2008) provided figures showing the complete polar lipid profiles. This information is of the highest importance when lipid profiles are composed mainly of unknown components, as is the case in chryseobacteria. When detailed comparative chromatographic results are missing from the description, the presence of the same unknown lipid in two strains can only be estimated by visual comparison of its chromatographic behaviour, together with the staining behaviour of the unknown lipid and in correlation with known lipids in the same chromatogram. Still, from the available data, it is possible to conclude that the polar lipid profile of K. koreensis KCTC 12107^T is quite similar to those of C. defluvii, C. gambrini, C. molle, C. pallidum and C. ureilyticum, indicating a close relatedness, which supports the inclusion of K. koreensis in the genus Chryseobacterium.

Fatty acid analysis was carried out as described previously (Kämpfer & Kroppenstedt, 1996; Kämpfer et al., 2003) on the eight isolates and on the type strains of C. haifense and K. koreensis. The fatty acid profiles of the eight strains studied were very similar to those of the type strains of C. haifense and K. koreensis (Table 3⇓) and in accordance with the fatty acid patterns reported for other Chryseobacterium species.

Hence, all data resulting from this polyphasic study demonstrate that the six clinical isolates NF 1366^T, NF 1101, NF 1107, NF 1123, NF 1413 and CCUG 15260 represent a novel species in the genus Chryseobacterium, for which the name Chryseobacterium anthropi sp. nov. is proposed.

K. koreensis was shown to be genotypically and phenotypically very similar to Chryseobacterium species. Its polar lipid profile exhibited the major characteristics shown for recently described Chryseobacterium species and the fatty acid profile of the type strain of K. koreensis was also very similar to those of Chryseobacterium species. The species C. anthropi sp. nov., C. haifense and K. koreensis cluster most closely according to 16S rRNA gene sequences and tDNA-PCR patterns and are negative for aesculin hydrolysis and ethylene glycol production, whereas these biochemical characteristics are positive for most other Chryseobacterium species. Because no genotypic or phenotypic differences could be found to justify the classification of K. koreensis in a separate genus, it is proposed to reclassify Kaistella koreensis in the genus Chryseobacterium as Chryseobacterium koreense comb. nov.