Genome-Based Taxonomic Classification of Bacteroidetes

1Bacteroidetes comprise bacteria widespread in the biosphere and isolated from many distinct habitats, including temperate, tropical and polar ecosystems (Krieg et al.,; Thomas et al., ). Bacteroides are anaerobic and mostly found in the gastrointestinal tract of animals and humans and, besides Firmicutes, even dominates the gut microflora of mammals (Smith et al.,; Ley et al.,; Thomas et al., ). Several Cytophagia were cultured from marine habitats, whereas Cytophaga was found in soils and Cyclobacteriaceae in (hyper-)saline environments only (Krieg et al., ). Flavobacteriaceae have colonized many different ecosystems such as soils, sediment, freshwater, brackish water, and seawater in temperate, tropical, and polar ecosystems (Bernardet, ). Some Flavobacteriaceae are even pathogenic for humans, other mammals, freshwater fish or marine fish. However, Blattabacteriaceae are endosymbiotic bacteria (e.g., in termites and cockroaches), whereas Cryomorphaceae mostly live in cold, marine environments (Bowman et al., ). The wide variety of habitats reflects the importance of Bacteroidetes in biogeochemical processes. For instance, aquatic, terrestrial and gut Bacteroidetes are well known for their functional specialization on the decomposition of peptides and polysaccharides (Kirchman,; Bowman,; Thomas et al.,; Fernández-Gómez et al., ). This feature is accompanied by a great number and diversity of carbohydrate-active enzymes (Cantarel et al., ) in Bacteroidetes genomes (Fernández-Gómez et al., ). The corresponding genes cluster together with TonB-dependent transporters in polysaccharide-utilization loci (Martens et al.,; Sonnenburg et al.,; Fernández-Gómez et al., ). Genome sequences are expected to support investigating the evolutionary relationship of gut Bacteroidetes and the diet of their hosts, facilitated by lateral gene (carbohydrate-active enzymes) and gene cluster (polysaccharide utilization loci) transfer between environmental and gut Bacteroidetes (Thomas et al., ).

2Bacteroidetes are Gram-stain-negative, chemo-organotrophic rods that do not form endospores and are either non-motile or motile by gliding (Woese,; Paster et al., ). Before the phylum was called Bacteroidetes (Krieg et al., ), it had been referred to as Cytophaga-Flavobacteria-Bacteroides (Paster et al.,; Woese, ). The phylum comprises the classes Bacteroidia, Cytophagia, Flavobacteriia, and Sphingobacteriia (Krieg et al., ). Recently, the name Bacteroidaeota was proposed for this phylum by including the rank phylum in the International Code of Nomenclature of Prokaryotes (Oren et al., ). As Cytophaga, Flexibacter, and Flavobacterium have many phenotypic characteristics in common, their differentiation used to be based on the presence or absence of gliding motility (Bernardet et al., ). However, gliding motility is a common feature of many Bacteroidetes genera (McBride and Zhu, ). Cytophaga and Flexibacter were also delineated based on cell morphology, the G+C content as well as the habitats they were isolated from Reichenbach. The anaerobic Bacteroidia used to be considered separate from the aerobic groups such as Flavobacteriia and Cytophagia, but 16S rRNA gene sequencing clarified their interconnections (Paster et al.,; Weisburg et al.,; Woese, ). Despite its usefulness in resolving such taxonomic questions, the 16S rRNA gene contains only a limited number of characters and thus usually yields only partially resolved phylogenies, i.e., trees with many statistically unsupported branches (Klenk and Göker,; Breider et al., ). Recently, Munoz et al. revised the phylogeny of Bacteroidetes and removed the incertae sedis taxa from the Balneola group and Rhodothermaceae (Bacteroidetes Order II. incertae sedis) from the Bacteroidetes as the novel phylum Rhodothermaeota. Nevertheless, many unsatisfactory aspects of Bacteroidetes classification might still persist.

3Indeed, only monophyletic taxa can be accepted in a taxonomic classification because its purpose is to summarize the phylogeny of the classified organisms (Hennig,; Wiley and Lieberman, ), and genome-scale data are more promising than single genes, or multi-locus sequence analysis restricted to a low number of genes, to identify monophyletic and non-monophyletic groups with high confidence (Klenk and Göker, ). The phenomenal increase in the number of publicly available whole-genome sequences further demands a genome-based classification system in today's genomic era.

4The genomic G+C content, i.e., the proportion of cytosines and guanines among all nucleotides in the genome, is one of the most frequently used taxonomic markers in microbiology (Mesbah et al.,; Rosselló-Mora and Amann, ). Within Bacteroidetes, it is strongly recommended to include the G+C content especially when describing every species of Flavobacteriaceae (Bernardet et al., ). The rapid progress in sequencing technology (Liu et al.,; Mavromatis et al., ) allows not only for inferring genome-scale phylogenies (Klenk and Göker,; Meier-Kolthoff et al., ) but also for replacing traditional methods that indirectly determine the G+C content (Mesbah et al.,; Moreira et al., ) by calculating it directly from highly accurate genome sequences. For this reason, literature claims that the variation of the G+C content within bacterial species is at most 3 mol% (Mesbah et al., ) or even up to 5 mol% (Rosselló-Mora and Amann, ) can be attributed to the imprecision of traditional methods (Meier-Kolthoff et al., ). On the other hand, within-species variation is at most 1% when both species boundaries (Meier-Kolthoff et al., ) and G+C contents are determined from genome sequences (Meier-Kolthoff et al., ). These inconsistencies call for correcting species descriptions that include a conventionally determined G+C content value that differs by more than 1% from the value calculated from the genome sequence of the type strain (Meier-Kolthoff et al.,; Riedel et al., ). Apparently the same holds if a range of G+C values was provided whose lower or upper bound deviates by more than 1% from the directly calculated G+C value, but a species description should also be restricted if a range of G+C values was provided that, unrealistically, exceeds 1%. Taxa of higher rank should accordingly be emended if their presumed range of G+C values turns out to be in conflict with the information from genome sequences (Scheuner et al., ).

5The Genomic Encyclopedia of Bacteria and Archaea (GEBA) pilot phase as well as the One Thousand Microbial Genomes phase 1 (KMG-1) projects (Kyrpides et al., ) are ideal data sources for genome-scale taxonomic reasoning because these projects aimed at filling the genomic gaps in the bacterial and archaeal branches of the tree of life (Göker and Klenk, ) and included only type strains with a certified origin from a culture collection and thus a verifiable history. Using the Bacteroidetes genome sequences from these projects, we here address the following questions: (i) What is the relationship between the phylogenomic trees and the proposed taxonomic classifications or the 16S rRNA gene phylogenies? (ii) Which taxa need to be revised because they are evidently non-monophyletic? (iii) Which taxon descriptions that lack G+C values should be augmented with information from genome sequences? (iv) Which taxon descriptions deviate from G+C content values calculated from genome sequences and should now be emended accordingly? (v) Does the correction of G+C values improve their fit to the phylogeny?

6The Bacteroidetes (ingroup), Chlorobi, Planctomycetes, and Verrucomicrobia (outgroup) type-strain genomes originating from the GEBA pilot phase and the KMG-1 (Kyrpides et al., ) project were downloaded from IMG (Markowitz et al., ) and augmented with additional type-strain genomes of taxonomic interest deposited in INSDC. The complete list is found in Supplementary Table 1. Genome-scale phylogenies were inferred from whole proteomes using the high-throughput version (Meier-Kolthoff et al., ) of the Genome BLAST Distance Phylogeny (GBDP) approach (Henz et al.,; Auch et al., ). BLAST+ (v2.2.30) (Camacho et al., ) was run in BLASTP mode with default parameters except for an e-value filter of 10−8 (Meier-Kolthoff et al., ). The greedy-with-trimming algorithm, which conducts a correction for non-orthologous hits, was applied in conjunction with formula d5, which relates the (weighted) number of identities within BLAST hits (high-scoring segment pairs) to the overall length of these hits and thus is unaffected by incomplete genome sequencing, and subjected to 100 pseudo-bootstrap replicates (Meier-Kolthoff et al.,, ). Phylogenetic trees were inferred from the original and pseudo-bootstrapped intergenomic distance matrices using FastME (Lefort et al., ), and tree and support values visualized using ITOL (Letunic and Bork, ). Where species affiliations had to be clarified, digital DNA:DNA hybridization was conducted with the recommended settings of the Genome-To-Genome Distance Calculator (GGDC) version 2.1 (Meier-Kolthoff et al., ).

7Comprehensive, aligned, near full-length 16S rRNA gene data for Bacteroidetes and the outgroup phyla were taken from version s123 of the All-Species Living Tree Project (LTP) (Yarza et al., ). Sequences of species of interest missing from LTP s123 (see Supplementary File 2) were added to the alignment using POA (Lee et al., ). Trees were inferred from the alignment with RAxML (Stamatakis, ) under the maximum-likelihood (ML) criterion in conjunction with the GTR-CAT model, fast bootstrapping, bootstopping (Pattengale et al., ) and subsequent search for the best tree, and with TNT (Goloboff et al., ) under the maximum-parsimony criterion (MP); here, new-technology search for the best tree was conducted as well as 1000 bootstrapping replicates in conjunction with tree-bisection-and-reconnection branch swapping and one random sequence addition replicate per bootstrap replicate. Further ML and MP trees were inferred in the same way but using the branches of the GBDP tree with ≥95% support as backbone constraint. Finally, 16S rRNA gene trees reduced to genome-sequenced strains were inferred.

8All trees were compared to the current classification used in LTP version s123, which was cleaned from inconsistencies, such as mismatches between species and genus names, and subsequently modified manually in the case of taxonomic arrangements that could not be confirmed in the literature, of validly published synonyms that were in better agreement to the phylogenies, and of missing taxa such as the family for Saccharicrinis (Yang et al., ). The full list of changes is described in Supplementary File 2. Whether taxa were monophyletic, paraphyletic or polyphyletic (Farris,; Wood, ) was determined using program code developed at DSMZ, which reports, in the case of taxa appearing non-monophyletic in a phylogeny, the highest support of all branches that conflict with the monophyly of that taxon as support against it (that can be displayed as negative support). In a rooted tree, these conflicting branches are the ones connected to subtrees that contain some but not all representatives of the taxon, as well as representatives of other taxa.

9The (trivial) calculation of the G+C content from genome sequences was done as in a previous study (Meier-Kolthoff et al., ) and by the GGDC server version 2.1. The changes in the fit between phylogeny and G+C content data when switching from conventionally calculated values to G+C counts from genome sequences was assessed by calculating their MP score with TNT (Goloboff et al., ), which allows for treating continuous characters as such (Goloboff et al., ). The data were rescaled to fit in the range between 0 and 65 as necessary for TNT and exported using the opm package (Vaas et al., ) for R (R Development Core Team, ); species without a literature G+C value were deactivated. To address phylogenetic uncertainty, the difference was re-calculated for each pseudo-bootstrap tree, and the resulting set of differences tested for its difference from zero using a t-test and Wilcoxon signed rank test as implemented in R. Because some G+C contents are given as ranges in the literature, we alternatively assessed minima, averages and maxima of the ranges, combined with the single G+C values from the other sources. For each species, only the most recent emendation that included a G+C content value was considered. Visualization was done with ggplot2 (Wickham, ). We further investigated the effect of incomplete genome sequencing on the calculation of G+C content values using the simulation techniques from our earlier study (Auch et al., ); details are provided in Supplementary File 3.

10Figure 1A shows the relationship between conventionally determined G+C content values and those calculated from the genome sequences; Supplementary Table 1 contains all collected data. A total of 161 species descriptions could be confirmed regarding the G+C content, for 45 it was obvious that they should be restricted because the provided G+C content range was unrealistically broad, 119 deviated by more than 1% from the genome sequence, and 24 lacked a G+C content range or value altogether. The MP scores of the conventionally determined and genome-sequence based G+C contents are shown in the Figure 1B. As confirmed by all tests conducted (α <0.001), it was obvious that the corrected G+C content values display a significantly better fit to the phylogeny than the original ones, and that this progress can be detected even though a large number of the G+C content values needed no correction. The results are in agreement with the simulations to assess the affect of incomplete genome sequencing on the calculation of the G+C content, which showed that the expected deviation from the real value is significantly below 0.1% for the given sequencing quality (Supplementary File 3).

11The underlying GDBP tree is shown in Figures 2–4 together with branch support values, taxonomic annotations and genomic G+C content indicators. Figure 5 compares the positive or negative support values for each taxon implicit in the GBDP (average support over all branches, 92.00%), 16S rRNA gene ML (69.47%), and 16S rRNA gene MP (67.25%) trees. The whole-genome tree supports more taxa than the 16S rRNA gene but also yields conflict in few cases where the 16S rRNA gene phylogeny is inconclusive, whereas several taxa are strongly supported as non-monophyletic by all approaches. However, in contrast to the classification, no well-supported (Taylor and Piel, ) discrepancies between the GBDP tree and the 16S rRNA gene trees were detected, as evident from the empty upper left and lower right corners in Figure 5. The unconstrained comprehensive 16S rRNA gene ML and MP trees (UCT) as well as the backbone-constrained ones (CCT) are included in Supplementary File 4. Whereas the majority of taxa appeared monophyletic in our analyses, a couple of discrepancies between the phylogenomic tree and the classification were observed, which we report herein in decreasing order of taxonomic rank involved, along with according suggestions for reclassifications.

12All phyla appeared as monophyletic in the GBDP tree (Figure 2). Given their separation from Bacteroidetes, the Balneolia and Rhodothermia could be assigned to a single phylum, Rhodothermaeota (Munoz et al., ). However, the 16S rRNA gene alone provides no support for their sister-group relationship (albeit no significant conflict), and the two groups strongly differ in their genomic G+C content (Figure 2), as previously noted (Urios et al., ); none of the remaining phyla show a comparable discrepancy in G+C values. Exactly the same reasons argue against an inclusion of the two groups in the phylum Chlorobi (Figure 2). Moreover, there are considerable phenotypic and habitat differences between the Balneolia and Rhodothermia (Table 1). The Balneolia are mesophilic, whereas Rhodothermia are mostly either strongly thermophilic (Alfredsson et al., ) or extremely halophilic (Antón et al., ). Carotenoids were only reported from Rhodothermia (Makhdoumi-Kakhki et al., ) but not from the Balneolia. For these reasons, the Balneolia and Rhodothermia should not only be removed from Bacteroidetes but also classified in two separate phyla, Rhodothermaeota (Munoz et al., ) and Balneolaeota (this study), based on the priority of the respective genus names. A comparable situation has been solved recently for the Ignavibacteriaceae (Iino et al., ) and Melioribacteraceae (Podosokorskaya et al., ). These taxa have been separated from the Chlorobi as a new phylum Ignavibacteria (Podosokorskaya et al., ) based on their physiological distinction and their large phylogenetic distance to Chlorobi.

13The Balneolaeota (Munoz et al., ) and Rhodothermaeota (as suggested here) differ from Bacteroidetes by a considerable number of phenotypic characters. The large amount of phospholipids [diphosphatidylglycerol (DPG), phosphatidylethanolamine (PE), phosphatidylglycerol (PG) and phosphatidylcholine (PC)] found in Rhodothermus (and its relatives) is unusual for aerobic Bacteroidetes (Nolan et al., ). Polar lipids of Bacteroides comprise largely PE and small amounts of PG and phosphatidylserine (Wardle et al., ), whereas polar lipids of the phylum Bacteroidetes usually comprise significant amounts of amide-linked lipids (aminolipids) rather than ester-linked polar lipids (Bernardet, ). The most significant difference between the Balneolaeota and Rhodothermaeota on the one hand and Bacteroidetes on the other hand can be attributed to the composition of fatty acids. Characteristic fatty acids of Bacteroidetes are 2-hydroxy and 3-hydroxy fatty acids, predominantly iso-C17:0 3-OH, iso-C15:0 3-OH and iso-C16:0 3-OH (Mayberry,; Krieg et al., ), whereas strains of the Balneolaeota and Rhodothermaeota do not possess 2-hydroxy and 3-hydroxy fatty acids (Table 1). Flagella in conjunction with motility were observed in Balneola vulgaris (Urios et al., ) and Salinibacter ruber (Antón et al., ) but are unusual in Bacteroidetes; conversely, neither gliding motility nor flexirubin, a pigment common in Bacteroidetes (Krieg et al., ), were reported for the Balneolaeota and Rhodothermaeota. We suggest according emendations of taxon descriptions.

14With the Balneolaeota and Rhodothermaeota removed from Bacteroidetes, the classes and orders of Bacteroidetes appeared as monophyletic in the GBDP phylogeny (Figure 2) with the sole exception of Chitinophagia and Chitinophagales, which were shown as paraphyletic due to the early branching of Haliscomenobacter, Phaeodactylibacter, and Lewinella (Saprospiraceae). Three branches with weak to moderate (79, 92, and 66%) support, respectively, would need to be wrong to obtain monophyletic Chitinophagia and Chitinophagales. The UCT and CCT show a distinct picture with a monophyletic Chitinophagia and Chitinophagales with 94–99% support under ML and <60–79% support under MP. Thus, regarding the monophyly criterion, it might or might not be adequate to place Saprospiraceae in Chitinophagales according to our analyses. The monophyly of the family itself, including the type genus Saprospira, is supported by the 16S rRNA gene trees with 95–100%. Thus, given its uncertain position relative to the remaining Chitinophagia and Chitinophagales, the taxonomic placement of Saprospiraceae should be reconsidered.

15The family Saprospiraceae has been proposed first in Bergey's manual (Garrity and Holt, ) and was suggested to represent a sister lineage, without confirmatory evidence, to the family Sphingobacteriaceae. Saprospiraceae was later on placed in Chitinophagales but without relevant branch support (Munoz et al., ). In contrast to both Sphingobacteriaceae and Chitinophagaceae, Saprospiraceae are long rods (up to 5 μm) that form long filaments (up to 500 μm) and do not possess sphingophospholipids. Moreover, the 16S rRNA gene trees indicate that the group comprising both Saprospiraceae and Chitinophagaceae (which is not even monophyletic in Figure 2) is quite divergent. Based on these results, we propose to classify the family Saprospiraceae into the new order Saprospirales, of the new class Saprospiria. Moreover, some noticeable phenotypical characteristics in agreement with the branching order (Figure 2, Supplementary File 4) call for a split of the family Saprospiraceae into three families (Saprospiraceae, Lewinellaceae, and Haliscomenobacteraceae). Lewinella, Saprospira and Aureispira exhibit gliding motility, but only Saprospira and Aureispira form helical filaments (Krieg et al., ). In contrast, Haliscomenobacter, Phaeodactylibacter, and Portibacter are non-motile straight filaments.

16A noticeable feature of the current classification of Bacteroidetes is that each class contains only a single order and thus the classification into orders provides no additional information. As this is not an issue of non-monophyly, we suggest addressing it once an even more comprehensive genome-scale phylogeny might help making more sense from these Linnaean ranks.

17The recently revised (Munoz et al., ) Rhodothermales families appeared as monophyletic (Figure 2) but turned out to differ significantly by some key physiological characteristics. Rhodothermus (Rhodothermaceae) tolerates up to 6% NaCl and 70°C (Table 1). In contrast, Salinibacter and Salisaeta (Salinibacteraceae) thrive in salt lakes and crystallizer ponds and grow at maximum temperatures of 50°C in medium with at least 5% NaCl and up to saturation (Table 1). Additionally, the polar lipids of Rhodothermus comprise mainly DPG and PE whereas Salinibacter, Salinivenus, and Salisaeta additionally contain PC, glycolipids and halocapnines, while Rubricoccus and Rubrivirga (Rubricoccaceae) contain PG, but no glycolipids (Table 1). Rhodothermus mainly displays saturated fatty acids, whereas Salinibacter, Salinivenus, and Salisaeta contain unsaturated and C18 fatty acids (Table 1). We thus suggest according emendations.

18The families within Sphingobacteriales appeared monophyletic throughout with maximum support in the GBDP tree (Figure 2). Within Cytophagales, the situation seems more complex. Cytophagaceae were not shown as monophyletic, since strong conflicting support was present regarding the positioning of Flexibacter and Microscilla, which are currently placed in that family (Nakagawa, ) but here Flexibacter appeared problematic per se and thus is discussed below in more detail.

19Flexithrix is placed apart from the remaining Flammeovirgaceae (Marivirga, Roseivirga) with 99% support in the GBDP analysis. Flexithrix and Rapidithrix formed a clade with reasonable bootstrap support (always >90%) in the UCT and CCT. Marivirga and Roseivirga did not form a clade either, with a conflicting branch supported by 94% in the GBDP tree. Because of their overall lower resolution, the 16S rRNA gene trees do not indicate in which of these three distinct groups of Flammeovirgaceae its type genus, Flammeovirga, is placed. For this reason, further revisions of Flammeovirgaceae have to be postponed until more genome sequences are available. Based on phylogenetic results (Nakagawa et al., ) and a polyphasic approach (Nedashkovskaya et al., ), the genus Marivirga was proposed to comprise of M. sericea and M. tractuosa (Nedashkovskaya et al., ). However, the physiological differentiation in that study only included Flammeovirgaceae genera (Fabibacter, Fulvivirga, Marinoscillum, Reichenbachiella, Roseivirga), and Thermonema lapsum (Flammeovirgaceae) was used as an outgroup for the 16S rRNA gene sequence-based phylogeny. A similar set of strains was investigated in the description of the two other Marivirga species, M. lumbricoides (Xu et al., ) and M. atlantica (Lin C.-Y. et al., ). Thus, the strain samplings in those earlier studies did not allow for detecting a closer affiliation of Marivirga to another family, which might explain the conflicting result obtained here.

20Within Bacteroidia, Porphyromonadaceae appeared paraphyletic with maximum support against their monophyly in the GBDP analysis (Figure 3), as all other genera were placed in clades distinct from the one harboring Porphyromonas. According to the GBDP phylogeny, three families should be separated from Porphyromonadaceae, each corresponding to a maximally support clade. A revision of the family is currently already conducted (Hugenholtz, pers. comm.), hence no taxonomic consequences will be undertaken in our study.

21The recent removal of Butyricimonas and Odoribacter from Porphyromonadaceae to place them in the new family Odoribacteraceae (Munoz et al., ) also requires some attention. The features of the genera that used to be included in Porphyromonadaceae have been reviewed in 2009 (Sakamoto et al., ); the statements below also extend to the more recently added genera Falsiporphyromonas (Wagener et al., ), Macellibacteroides (Jabari et al., ), and Petrimonas (Grabowski et al., ). (An issue with Falsiporphyromonas and Macellibacteroides that is beyond the scope of the current study is that these genera are nested with strong support within Porphyromonas and Parabacteroides, respectively; see Supplementary File 4.) Butyricimonas and Odoribacter (Hardham et al., ) differ from each other, but not clearly from the other genera, by their major metabolic end products; however, this feature does not unambiguously differentiate between the other genera either. Butyricimonas and Odoribacter differ from all other genera except Porphyromonas by containing iso-C15:0 as major fatty acid, with a low ratio of anteiso-C15:0 to iso-C15:0. Butyricimonas and Odoribacter differ from Falsiporphyromonas and Porphyromonas by their fermentative metabolism (Sakamoto et al., ) and partially by their lack of pigments (Nagai et al., ). To reflect this distribution of phenotypic features, we propose to emend the family Odoribacteraceae accordingly.

22Six genera were found to be non-monophyletic in the phylogenomic analysis. Pedobacter (Sphingobacteriaceae) appeared paraphyletic because Pseudopedobacter saltans was placed as sister group of Pedobacter glucosidilyticus with 100% support and Pedobacter oryzae as sister group of Arcticibacter svalbardensis and Mucilaginibacter paludis with <50% support (Figure 2). The UCT and CCT also showed Pedobacter distributed over several clades but without significant support against its monophyly. Due to the size of the genera Mucilaginibacter and Pedobacter we suggest sampling more genomes from these groups prior to drawing taxonomic conclusions.

23Flexibacter (Cytophagaceae) appeared paraphyletic with high confidence in the phylogenomic tree because the clade comprising F. litoralis and F. roseolus (separated by long branches) occurred in clades with 77 and 99% support together with other genera, to the exclusion of F. elegans, which formed the sister group of Microscilla marina with maximum support (Figure 2). A paraphyletic Flexibacter also occurred in the UCT and CCT, with strong support against its monophyly, as F. elegans was also highly supported as sister group of Microscilla marina. F. ruber formed the sister group of that clade without significant support, whereas Flexibacter flexilis Soriano 1945, the type species of the genus (Nakagawa, ), was indeed placed within the Cytophagaceae, with Arcicella, Flectobacillus, and Pseudarcicella as most closely related genera, without much support. Flexibacter litoralis, F. polymorphus, and F. roseolus formed a clade with reasonable bootstrap support (>90% in the CCT), which was placed in an uncertain position. Apart from tree topology, the branch lengths in the whole-genome and 16S rRNA gene tree indicated that the Flexibacter species are too divergent to be placed in a single genus. Additionally, Microscilla marina and F. elegans appear to be too divergent to be placed in a single genus (Figure 2).

24Since its description, eleven of 17 originally supposed Flexibacter species were reclassified into a variety of genera (Chitinophaga, Flavobacterium, Marivirga, Solitalea, and Tenacibaculum), but it was emphasized that Flexibacter is still heterogeneous and should be subdivided based on additional molecular taxonomic data (Nakagawa, ). Furthermore, this author mentioned that the type species, Flecibacter flexilis, is isolated from all other species of the genus and that Flexibacter should be restricted to it. The current sole distinction between the definitions of Microscilla and Flexibacter is their habitat (marine vs. terrestrial) (Reichenbach, ), whereas F. elegans showed the same morphology and physiology as M. marina in previous studies (Lewin, ); F. roseolus differed from M. marina regarding its pigmentation and F. ruber by its ability to digest starch. F. flexilis differs from M. marina and F. elegans regarding cell length and H2S production, whereas the other species differ from F. flexilis regarding pigmentation, cell length and in the case of F. roseolus also H2S production. Thus, the Flexibacter species are approximately equidistant from each other regarding morphology and physiology. This also holds for Flexibacter litoralis and F. polymorphus; whereas they form sister groups with high support, their level of divergence is higher than between many other pairs of genera, as obvious from the branch lengths in the trees (Figures 2, 3; Supplementary File 4), and their morphology and physiology also differs (Lewin,, ). Thus, based on phylogenetic position (Supplementary File 4) and phenotype, we propose to classify F. elegans, F. litoralis, F. polymorphus, F. roseolus, and F. ruber into the new genera Eisenibacter, Bernardetia, Garritya, Hugenholtzia, and Thermoflexibacter, respectively. Further, we propose the new family Microscillaceae to accomodate Bernardetia and Eisenibacter as well as the new family Bernardetiaceae to accommodate Bernardetia, Hugenholtzia and tentatively also Garritya.

25Within Prevotellaceae, Prevotella (Shah and Collins,; Willems and Collins,; Sakamoto and Ohkuma, ) appeared paraphyletic because Hallella seregens (Moore and Moore, ) was included within, placed as sister group of P. dentalis with strong support (Figure 3). This problem was also obvious in 16S rRNA gene data and was indeed noted two decades ago (Willems and Collins, ). These authors refrained from taxonomic consequences, however, because they observed a 16S rRNA gene similarity of 99.8% between the type strains of H. seregens and P. dentalis; hence it could not be excluded that the two species are heterotypic synonyms (Meier-Kolthoff et al., ). Furthermore, the G+C values calculated from the genomes of H. seregens and P. dentalis are almost identical, with 56.0 and 55.9%, respectively, even though this is not currently reflected in the species descriptions. Digital DNA:DNA hybridization between the genome sequences of H. seregens ATCC 51272T and P. dentalis DSM 3688T conducted with the recommended settings of the GGDC yielded 87.40 ± 2.36% DDH similarity and thus indicated identical species, even identical subspecies (Meier-Kolthoff et al., ). We conclude that H. seregens is a later heterotypic synonym of P. dentalis.

26Within Flavobacteriaceae, Chryseobacterium appeared paraphyletic because Epilithonimonas tenax was clearly nested within a well-supported branch of 21 Chryseobacterium species (Figure 4). Whereas support was maximal in the phylogenomic analysis, the UCT and CCT showed Epilithonimonas nested within a paraphyletic Chryseobacterium, too, but without any branch support. In the original description of Epilithonimonas (O'Sullivan et al., ) only 75% bootstrap support was obtained for the separation of Epilithonimonas from Chryseobacterium, based on a simplistic evolutionary model (Felsenstein, ). The most recent description of an Epilithonimonas species (Ge et al., ), still based on that model, showed moderate support for Epilithonimonas monophyly but none for Chryseobacterium monophyly; hence 16S rRNA gene analyses did not rule out that Epilithonimonas is nested within Chryseobacterium, as revealed by phylogenomic analysis. As evident from Figure 4, retaining Epilithonimonas by including certain Chryseobacterium species would require the establishment of a third genus for at least the upper branch including six Chryseobacterium species (C. koreense, C. antarcticum, C. solincola, C. palustre, C. jeonii, C. bovis) to obtain monophyletic groups. The solution to merge the two genera is thus more conservative. Moreover, the overall divergence of the clade comprising the two genera is lower than the divergence of other clades comprising only a single genus (Figures 2–4). Chryseobacterium has priority over Epilithonimonas (Vandamme et al., ). Chryseobacterium was recently emended because some summed features of MIDI-system fatty-acids profiles appeared to have frequently been misinterpreted in the literature, and a specific pattern of polar lipids remained mostly unnoticed (Montero-Calasanz et al., ). In detail, iso-C15:0 2-OH is present in Chryseobacterium as part of summed feature 3 or 4 instead of C16:1ω7t, C16:1ω6c and C16:1ω7c; iso-C17:1ω7c is present instead of iso-C17:1ω9c; and the major polar lipids contain three common unknown lipids and two common unknown aminolipids.

27Thus, the question arises whether the reported differences regarding the lipid profiles between Epilithonimonas and the description of Chryseobacterium can be traced back to misidentified fatty acids or were real differences. Table 2 shows the potentially conflicting phenotypic features. The only real difference in fatty-acid profiles appeared to be the presence of iso-C16:0 3-OH, which was not explicitly mentioned for the genus Chryseobacterium (Montero-Calasanz et al., ). However, it was reported for a variety of Chryseobacterium species (Hugo et al.,; Kämpfer et al.,,; Li et al.,; Kim et al.,; Shen et al.,; Young et al.,; Park et al.,; Behrendt et al.,,; Hantsis-Zacharov and Halpern,; Quan et al.,; Vaneechoutte et al.,; Szoboszlay et al.,; Ilardi et al.,; Benmalek et al.,; Pires et al.,; Wu et al., ) in concentrations up to 9% and thus does not constitute a real difference either. Additionally, anteiso-C15:0 as reported for Epilithonimonas is not mentioned in the description of Chryseobacterium but known to occur in the genus from traces to up to 8.2% (Montero-Calasanz et al., ). The polar-lipid patterns of Epilithonimonas also fit very well to that revision of Chryseobacterium (Table 2); the sole exception is E. psychrophila, for which only phosphatidylethanolamine and a single unknown lipid were reported; given its well-supported position within a core group of four Epilithonimonas species and the general faintness of the published thin-layer chromatogram (Ge et al., ), this can hardly be regarded as an argument against the unification of the genera. Moreover, the two genera have overlapping habitats and other interesting features in common, such as the production of indole-3-acetic acid for promoting plant growth (Montero-Calasanz et al.,; Hoang et al., ). Thus, the three physiological features that conflict between the original description of Chryseobacterium (Vandamme et al., ) and some or all of the Epilithonimonas species rather call for an emendation of Chryseobacterium.

28Muricauda (Flavobacteriaceae) appeared paraphyletic in the phylogenomic tree because Croceitalea dokdonensis was placed as sister group of M. lutaonensis. Support was weak (<60%), however, and in the UCT and CCT Croceitalea and Muricauda appear as separate groups with moderate support. Thus, even though in the description of the moderate thermophile M. lutaonens the genus Croceitalea was not considered (Arun et al., ), our analyses do not indicate a need to revise the classification of the two genera.

29Aequorivita (Flavobacteriaceae) also appeared paraphyletic in the phylogenomic tree because two Vitellibacter species were placed as sister group of Aequorivita sublithincola with 94% support. In the UCT and CCT the clade comprising both genera is maximally supported, but within the clade the two genera are intermixed without much support. The genomic divergence of the clade comprising both genera appears as lower than the divergences of many clades that contain only a single genus (Figure 3). When the genus Vitellibacter was introduced (Nedashkovskaya et al., ), type strains from the previously suggested genus Aequorivita (Bowman and Nichols, ) were not considered. The phenotypic features reported in the literature (Bowman and Nichols,; Nedashkovskaya et al.,; Bowman,; Park et al.,,; Kim et al.,; Liu et al.,; Lin S. Y. et al.,; Rajasabapathy et al., ) for the species of two genera considerably overlap. The only stable difference is whether the cells are oxidase-positive or negative, but this alone can hardly justify two separate genera. For this reason, we propose to place Vitellibacter in Aequorivita and to emend the description of Aequorivita accordingly.

30Our phylogenetic analysis revealed much agreement between genome-scale data and the current classification of Bacteroidetes, particularly due to the most recent revision (Munoz et al., ), but also a number of instances which still call for reclassifications or emendations of taxon descriptions. Prokaryotic taxonomy is not “written by taxonomists for taxonomists” (Cowan, ) but directly influences all microbiological disciplines. In microbial ecology, the name of a given taxon gets linked to observed characteristics of its described representatives and, further, to its inferred function in the environment. For example, in the Bacteroidetes-Chlorobi “superphylum,” Chlorobi (Garrity and Holt, ), now Chlorobaeota (Oren et al., ), are known as the obligate anaerobic green sulfur bacteria, whereas Bacteroidetes, now Bacteroidaeota (Oren et al., ), are known as polymer-decomposing microorganisms in the environment and the intestinal tract (Thomas et al., ). Currently Halomonadaceae, some Bacteroidetes (Rhodothermaceae, Salinibacter) and Chlostridia are considered as halotolerant or halophilic Bacteria (Oren,,; Bakermans, ). The reclassifications of the Rhodothermia and the Balneolia into separated phyla suggest that some Bacteroidetes remain thermotolerant (e.g., Schleiferia and Thermonema, growth at 30–60°C) or halotolerant [e.g., Anaerophaga, Fabibacter, and Dyadobacter, growth at 1–12% NaCl (w/v)] (Krieg et al., ), but not thermophilic (Rhodothermus, growth at 50–70°C) or halophilic [Salinibacter, growth at 15–30% NaCl (w/v)].

31In other areas such as medicine and industry, the risk classification of a prokaryotic species to cause infectious diseases is based on its taxonomic classification (ABAS,; ABSA, ). Given the peculiar ecological preferences of particularly Rhodothermaeota and Balneolaeota, but also of other groups reclassified here, it is likely that ecological studies using metagenomic binning or similar techniques will benefit from the further improved classification. Studies that do not distinguish between Rhodothermaeota, Balneolaeota, and Bacteroidetes may insufficiently describe microbial compositions of environmental habitats and perhaps make misleading assumptions on environmental conditions (Tsiamis et al.,; Baati et al.,; Çınar and Mutlu, ). A more profound effect of the proposed reclassifications might be on a potentially misleading proposal of an origin for horizontal-gene-transfer (Nelson-Sathi et al., ). In contrast, the CARD-FISH probes CF968 (Acinas et al., ) and CFB319 (Manz et al., ) will not hybridize with 16S rRNA gene sequences of Rhodothermaeota and the Balneolaeota (both probes have two mismatches), and thus environmental studies using this probe are not influenced by the mentioned reclassifications.

32Our analysis has shown that a large proportion of G+C content values in Bacteroidetes species descriptions needs to be corrected. The according emendations proposed below are numerous but by no means excessive. Indeed, we do not propose to correct all G+C content values whose precision or accuracy could be improved but only those that are too imprecise or inaccurate given that the within-species deviation in G+C content is at most 1% (Meier-Kolthoff et al., ). G+C content values provided in species descriptions that deviate more than 1% from the value calculated from the genome sequence, or display a range of more than 1%, obscure species affiliations and should be corrected (Riedel et al., ). Emendations of higher taxa regarding their G+C content have also been conducted (Scheuner et al., ) but we abstain from accordingly redefining Bacteroidetes genera here because we opine that more type-strain genome sequences per genus would be needed. Moreover, the emendation of Flavobacterium suncheonense has recently been conducted (Panschin et al., ) and needs not be considered here. For the emendations conducted below we round all G+C content values to a single decimal place to accommodate for incompletely sequenced genomes (see Supplementary File 3 for details).

33Whereas conflict with the monophyly of several taxa was obvious in our genome-scale analysis, cases where a taxon was strongly supported in the proteome tree and strongly in conflict with the 16S rRNA gene, or vice versa, were not observed (Figure 5). This is crucial since the promise of phylogenomics is to yield more strongly resolved trees (which fully succeeded in this study), but this might in theory also increase the topological conflict between analyses based on distinct large sets of characters or distinct phylogenetic methods (Jeffroy et al.,; Klenk and Göker, ). For instance, horizontal gene transfer is known as a cause of topological conflict between analyses of single genes and has even been be used to argue against hierarchical classification (Bapteste and Boucher,; Klenk and Göker, ). However, adding more genes, up to virtually all genes available in a genome, increases support in phylogenomic analyses (Breider et al., ), indicating that there is a strong hierarchical signal. In contrast, selection of few genes (or any number of characters not on the same order of magnitude as the overall number of genomic characters) can hardly be called genome-scale and also depends on assumptions about the relative suitability of genes for the analysis (Lienau and DeSalle,; Klenk and Göker, ). Whole-genome methods such as GBDP avoid this issue and have the potential to yield a truly genome-based classification, but then the question arises how to avoid overestimating phylogenetic confidence (Taylor and Piel, ). Bootstrapping entire genes instead of single alignment positions was suggested to reduce conflict and provide more realistic support values in phylogenomic analysis (Siddall, ). GBDP pseudo-bootstrapping in conjunction with the greedy-with-trimming algorithm (Meier-Kolthoff et al., ) is indeed conceptually closer to this partition bootstrap than to standard bootstrapping. Compared to analyses of concatenated alignments of orthologs, GBDP yielded the same topology but slightly lower support values (Riley et al., ). This might explain why we did not identify significant conflict with the 16S rRNA gene in this study.

34We neither identified any significant conflict between phenotypic information and reclassifications suggested by genome-scale trees; rather, these called for a distinct interpretation of the available information, or highlighted known problems that just had not yet been taxonomically fixed. Many of the taxonomic problems had also been visible in the 16S rRNA gene sequence analysis in the past, provided a sufficient taxon sampling had been used. For instance, sampling Chlorobi was also necessary in previous (Munoz et al., ) and the current phylogenetic analyses to fully recognize the problematic placement of the Balneola clade and Rhodothermaceae. In other cases, missing genomes did not hinder recognizing taxonomic problems. For instance, whereas the genome sequence of the type strain of the type species of Flexibacter was not available to us, the need to split the genus into several genera is already visible in the unconstrained 16S rRNA gene analysis. Moreover, applying a backbone constraint is a valuable means for augmenting comprehensible sampled single-gene data with information from analyses of more genes but fewer organisms (Liu et al., ).

35The high resolution provided by whole-genome phylogenies and the apparent lack of conflicts with data from other sources renders genome-based approaches quite promising for shaping the future of Bacteroidetes taxonomy. This holds even though the present study led to a number of changes in taxonomic opinion, listed below in the order of decreasing Linnaean ranks and with proposed names first, followed by the emendations. We additionally propose to rephrase the “Bacteroidetes-Chlorobi superphylum” to “Bacteroidaeota-Rhodothermaeota-Balneolaeota-Chlorobaeota (BRBC) superphylum,” considering the recent proposal to include the phylum rank in taxonomy (Oren et al., ), and considering the new phyla Rhodothermaeota and Balneolaeota as part of the former “Bacteroidetes-Chlorobi superphylum” (Figure 2).

36Bal.ne.o.lae.o'ta (N.L. fem. n. Balneola, type genus of the type order of the phylum, -aeota ending to denote phylum; N.L. neut. pl. n. Balneolaeeota, the phylum of the class Balneolia).

37The description is the same as for the class Balneolia.

38This phylum is described on the basis of 16S rRNA gene and whole-genome phylogenetic analysis. The type (and currently sole) order of the phylum is Balneolales. The phylum currently contains a sole class, Balneolia.

39The description is the same as given by Munoz et al. with the following modification. Cells are aerobic. The phylum has been additionally circumscribed on the basis of whole-genome phylogenetic analysis.

40This change was necessary due to the removal of Balneolia from the phylum Rhodothermaeota.

41The description is the same as given by Munoz et al. with the following modification. The class Balneolia is part of the phylum Balneolaeota and has been additionally circumscribed on the basis of whole-genome phylogenetic analysis.

42This change was necessary due to the removal of Balneolia from the phylum Rhodothermaeota. The description is the same as for the order Balneolales.

43The description is the same as given by Nakagawa with the following modification. Flagella are not found.

44This change was necessary due to the removal of Balneola from Cytophagia.

45Sa.pros.pi'ri.a (N.L. fem. n. Saprospira, type genus of the type order of the class, -ia ending to denote a class; N.L. fem. pl. n. Saprospiria, the class of the order Saprospirales).

46The description is the same as for the order Saprospirales.

47This class is described on the basis of 16S rRNA gene and whole-genome phylogenetic analysis. The type and currently sole order of the class is Saprospirales.

48Sa.pro.spi.ra'les (N.L. fem. n. Saprospira, type genus of the order, suff. -ales ending to denote order; N.L. fem. pl. n. Saprospirales, the order of the genus Saprospira).

49The description is the same as for the family Saprospiraceae.

50This order is described on the basis of 16S rRNA gene and whole-genome phylogenetic analysis. The type genus is Saprospira. The currently sole family of the order is Saprospiraceae.

51The description is as given by Munoz et al., with the following changes. Currently encompasses only the family Chitinophagaceae. This change was necessary due to the removal of Saprospiraceae from the order Chitinophagales.

52Le.wi.nel.la'ce.ae (L. fem. n. Lewinella type genus of the family; -aceae ending to denote a family; N.L. fem. pl. n. Lewinellaceae the Lewinella family).

53Cells are ensheathed, Gram-stain negative flexible rods (up to 3 μm) that form long filaments (up to 150 μm) and are motile by gliding. Typical fatty acids are iso-C15:0, iso-C15:1, iso-C17:0 3-OH and summed feature 3 (either C16:1 ω7c/C16:1 ω6c or C16:1 ω7c/iso-C15:0 2-OH). Seawater is required for growth. Flexirubin-type pigments are not produced. Carotenoid-type pigments are produced. The respiratory quinone is MK-7. The genomic G+C content is around 45–53%.

54This family belongs to the phylum Bacteroidetes, order Saprospirales ord. nov., class Saprospiria class. nov., and currently comprises only the type genus, Lewinella.

55Ha.lis.co.me.no.bac.te.ra'ce.ae (L. fem. n. Haliscomenobacter type genus of the family; -aceae ending to denote a family; N.L. fem. pl. n. Haliscomenobacteraceae the Haliscomenobacter family).

56Cells are Gram-stain negative, non-motile long rods (up to 5 μm) that form long needle-like filaments (up to 300 μm). Some are enclosed by a narrow, hardly visible hyaline sheath (Haliscomenobacter). Typical fatty acids are iso-C15:0, summed feature 3 (either C16:1 ω7c/C16:1 ω6c or C16:1 ω7c/iso-C15:0 2-OH) and either iso-C17:0 3-OH or iso-C15:0 3-OH. Flexirubin-type pigments are not produced. Carotenoid-type pigments are produced. The respiratory quinone is MK-7. Cells are oxidase-positive. The genomic G+C content is around 47–54%.

57This family belongs to the phylum Bacteroidetes, order Saprospirales ord. nov., class Saprospiria class. nov., and currently comprises the genera Haliscomenobacter (the type genus), Phaeodactylibacter and Portibacter.

58Mi.cros.cil.la'ce.ae (L. fem. n. Microscilla type genus of the family; -aceae ending to denote a family; N.L. fem. pl. n. Microscillaceae the Microscilla family).

59Cells are Gram-stain negative, chemoorganotrophs and strict aerobes. Cells are flexible threads motile by gliding. Gelatin is degraded, but not agar, alginate, or carboxymethyl cellulose. The major respiratory quinone is MK-7. The genomic G+C content is around 40–50%.

60This family belongs to the phylum Bacteroidetes, order Cytophagales, class Cytophagia, and comprises currently the genera Microscilla (the type genus) and Eisenibacter gen. nov.

61Ber.nar.de.ti.a'ce.ae (L. fem. n. Bernardetia type genus of the family; -aceae ending to denote a family; N.L. fem. pl. n. Bernardetiaceae the Bernardetia family).

62Cells are Gram-stain negative, chemoorganotrophs and strict aerobes that are motile by gliding. Threads are formed, with and without cross-walls. Carotenoids, saproxanthin or flexixanthin, are produced. Oxidase activity is present, catalase activity is absent. Gelatin is hydrolyzed, but neither cellulose nor agar. The major respiratory quinone is MK-7. The genomic G+C content is 29–38%.

63This family belongs to the phylum Bacteroidetes, order Cytophagales, class Cytophagia, and currently comprises the genera Bernardetia (the type genus), Hugenholtzia, and tentatively also Garritya.

64The description is the same as given by Munoz et al. with the following modification.

65Cells are non-motile or motile by means of flagella. The dominant fatty acids are iso-C15:0 and other non-hydroxy branched-chain fatty acids. Major polar lipids are diphosphatidylglycerol, phosphatidylethanolamine and either phosphatidylglycerol (Balneola, Gracilimonas) or phosphatidylcholine (Aliifodinibius, Fodinibius). The major menaquinone is MK-7. The genomic G+C content varies around 45%.

66The description is the same as given by Munoz et al. with the following modification.

67Major polar lipids are diphosphatidylglycerol (cardiolipid) and diphosphatidylcholin. Some species possess halocapnines. The major fatty acids are iso-C15:0, C18:1 ω7c, summed feature 3 (C16:1 ω6c and/or C16:1 ω7c).

68The description is the same as given by Ludwig et al. with the following modification.

69Major polar lipids are diphosphatidylglycerol, phosphatidylethanolamine and phosphatidylglycerol. The major fatty acids are iso-C16:0, iso-C17:0, anteiso-C17:0. The genomic G+C content varies around 65%.

70The description is the same as given by Munoz et al. with the following modification.

71Cells are non-motile. Metabolism fermentative, major end products are diverse organic acids. Major menaquinones are MK-9 and MK-10. Major fatty acid iso-C15:0, with a low ratio of anteiso-C15:0 to iso-C15:0. The genomic G+C content varies around 40–50%.

72The description is as given by Krieg et al., with the following changes.

73Cells are long rods (up to 3.5 μm) that form long helical filaments (up to 500 μm) and are motile by gliding. NaCl is required for growth and some can be tolerate NaCl at a concentration of up to 9% (w/v). Cytochrome oxidase and catalase activities are variable. Flexirubin-type pigments are not produced. Carotenoid-type pigments are produced. The respiratory quinone is MK-7. The genomic G+C content is around 33–48%.

74This family belongs to the phylum Bacteroidetes, order Saprospirales ord. nov., class Saprospiria class. nov., and currently comprises the genera Saprospira (the type genus) and Aureispira.

75Ber.nar.de'ti.a (N.L. masc. n. Bernardet named after Jean-François Bernardet, researcher at INRA Research Center, Jouy-en-Josas, France, and chairman of ICSP subcommittee on the taxonomy of aerobic Bacteroidetes; N.L. fem. n. Bernardetia a genus named after Jean-François Bernardet).

76On the basis of 16S rRNA gene sequence analysis, the genus represents a separate branch within the order Cytophagales. With moderate support it forms a clade together with Hugenholtzia gen. nov. and perhaps also Garritya gen. nov. but the three type species, respectively, are comparatively distant from each other in terms of sequence divergence, morphology and physiology. The type species of the genus is Bernardetia litoralis comb. nov.

77B. li.to.ra'lis (L. masc. adj. litoralis, belonging to the sea shore).

78Basonym: Flexibacter litoralis Lewin 1969

79The description is the same as for Flexibacter litoralis (Lewin, ). The type strain is ATCC 23117 = DSM 6794.

80Gar.ri'ty.a (N.L. masc. n. Garrity named after George M. Garrity, professor at Michigan State University, East Lansing, MI, USA, the former editor of Bergey's manual and current chief editor of Standards in Genomic Sciences; N.L. fem. n. Garritya a genus named after George M. Garrity).

81The description is the same as for the species Garritya polymorpha, comb. nov., as its known features, while scarce, already differentiate at the genus level.

82On the basis of 16S rRNA gene sequence and phylogenomic analysis, the genus represents a separate branch within the order Cytophagales with an uncertain affiliation to a family. It potentially forms a clade together with Bernardetia gen. nov. and Hugenholtzia gen. nov. but the three type species, respectively, are comparatively distant from each other in terms of sequence divergence, morphology and physiology. The type (and currently sole) species of the genus is Garritya polymorpha.

83G. po.ly.mor'pha (N.L. fem. adj. polymorpha, variable in form).

84Basonym: Flexibacter polymorphus Lewin 1974

85The description is the same as for Flexibacter polymorphus (Lewin, ). The type strain is ATCC 27820 = DSM 9678.

86Hu.gen.hol'tzi.a (N.L. masc. n. Hugenholtz named after Philip Hugenholtz, Professor at University of Queensland, Brisbane, Qld, Australia, who played a decisive role in starting the GEBA project; N.L. fem. n. Hugenholtzia a genus named after Philip Hugenholtz).

87The description is the same as for the species Hugenholtzia roseola, comb. nov., as its known features, while scarce, already differentiate at the genus level.

88On the basis of 16S rRNA gene sequence and phylogenomic analysis, the genus represents a separate branch within the order Cytophagales. With moderate support it forms a clade together with Bernardetia gen. nov. and perhaps also Garritya gen. nov. but the three type species, respectively, are comparatively distant from each other in terms of sequence divergence, morphology and physiology. The type species of the genus is Hugenholtzia roseola.

89H. ro.se'o.la (N.L. fem. dim. adj. roseola, with a rosy shading).

90Basonym: Flexibacter roseolus Lewin 1969

91The description is the same as for Flexibacter roseolus (Lewin, ) with the following restriction. The genomic G+C content is 42.2%. The type strain is ATCC 23088 = DSM 9546.

92Ther.mo.fle.xi.bac'ter (Gr. adj. thermos hot; L. part. adj. flexus bent, winding; N.L. masc. n. bacter from Gr. neut. n. baktron little stick or rod; N.L. masc. n. Thermoflexibacter a thermophilic flexible rod. The type strain GEY was isolated from a Geysir in Iceland, can grow in freshwater medium of up to 45°C and forms very long threads of more than 50 μm length.

93The description is the same as for the species Thermoflexibacter ruber, comb. nov., as its known features, while scarce, already differentiate at the genus level.

94On the basis of 16S rRNA gene sequence analysis as well as previously published physiological and morphological data, the genus represents a branch of uncertain affiliation within the order Cytophagales. The 16S rRNA gene quite weakly supports a sister-group relationship of this genus and the new family Microscillaceae. The type (and currently sole) species of the genus is Thermoflexibacter ruber.

95T. ru'bra (L. fem. adj. rubra, red).

96Basonym: Flexibacter ruber Lewin 1969

97The description is the same as for Flexibacter ruber (Lewin, ). The type strain is GEY = ATCC 23103 = DSM 9560.

98Ei.se.ni.bac'ter (N.L. masc. n. Eisen named after Jonathan A. Eisen, professor at University of California, Davis, CA, USA, who played a decisive role in starting the GEBA project; N.L. masc. n. bacter a rod; N.L. masc. n. Eisenibacter a rod named after Jonathan A. Eisen).

99The description is the same as for the species Eisenibacter elegans, comb. nov., as its known features, while scarce, already differentiate at the genus level.

100On the basis of 16S rRNA gene sequence analysis as well as previously published physiological and morphological data, the genus represents a separate branch within the order Cytophagales. The 16S rRNA gene strongly supports a sister-group relationship of this genus and Microscilla. The type (and currently sole) species of the genus is Eisenibacter elegans.

101E. e'le.gans (L. masc. adj. elegans refined, fashionable, elegant).

102Basonym: Flexibacter elegans (ex Lewin 1969, non Soriano, 1945) Reichenbach 1989b, 2067AL

103The description is the same as for Flexibacter elegans (Reichenbach, ). The type strain is ATCC 23112 = DSM 3317 = JCM 21159 = LMG 10750 = NBRC 15055 = Lewin NZ-1.

104The description is as given by Park et al. with the following modifications. Gliding motility is present or absent, flexirubin production is variable, oxidase reaction is positive or negative. The genomic G+C content varies around 37%.

105This change was necessary due to the inclusion of the Vitellibacter species in Aequorivita.

106A. aes.tu.a'ri.i (L. gen. n. aestuarii, of a tidal flat).

107Basonym: Vitellibacter aestuarii Kim et al. 2010

108The description is the same as for Vitellibacter aestuarii (Kim et al., ). The type strain is JC2436 = IMSNU 14137 = KACC 13727.

109A. e.chi.no.i.de.o'rum (N.L. gen. n. echinoideorum, of Echinoidea, sea urchins).

110Basonym: Vitellibacter echinoideorum Lin et al. 2015

111The description is the same as for Vitellibacter echinoideorum (Lin S. Y. et al., ). The type strain is CC-CZW007 = BCRC 80886 = JCM 30378.

112A. ni.o.nen'sis (N.L. adj. nionensis, derived from the acronym of the National Institute of Oceanography, NIO).

113Basonym: Vitellibacter nionensis Rajasabapathy et al., 2015

114The description is the same as for Vitellibacter nionensis (Rajasabapathy et al., ). The type strain is VBW088 = KCTC 32420 = MCC 2354.

115A. soe.sok.ka.ken'sis (N.L. adj. soesokkakensis, pertaining to the Soesokkak area)

116Basonym: Vitellibacter soesokkakensis Park et al., 2014

117The description is the same as for Vitellibacter soesokkakensis (Park et al., ) with the following modification. The genomic G+C content of the type strain is 37.8%. The type strain is RSSK-12 = KCTC 32536 = CECT 8398.

118A. vla.di.vos.to.ken'sis (N.L. adj. vladivostokensis, pertaining to the city of Vladivostok)

119Basonym: Vitellibacter vladivostokensis Nedashkovskaya et al. 2003

120The description is the same as for Vitellibacter vladivostokensis (Nedashkovskaya et al., ). The type strain is KMM 3516 = NBRC 16718.

121The description is the same as given by Wang et al. with the following modification. The genus Aliifodinibius is a member of the phylum Balneolaeota.

122This change was necessary due to the removal of Aliifodinibius from Rhodothermaeota.

123The description is as given by Montero-Calasanz et al. with the following modification. Almost all strains grow at 30°C. Most, but not all, strains oxidize glycerol and trehalose.

124On the basis of phylogenomic analysis and a re-assessment of 16S rRNA gene sequence analyses and phenotypic features published earlier, the genus Epilithonimonas should be included in Chryseobacterium.

125C. gin.sen.gi.ter'rae (N.L. n. ginsengum ginseng; L. n. terra soil; N.L. gen. n. ginsengiterrae of soil from a ginseng field).

126Basonym: Epilithonimonas ginsengisoli Hoang et al. 2015 (the name Chryseobacterium ginsengisoli has already been validly published, hence a new epithet must be chosen to avoid homonyms)

127The description is the same as for Epilithonimonas ginsengisoli (Hoang et al., ) with the following modification. Summed feature 3 should be interpreted as iso-C15:0 2-OH (Montero-Calasanz et al., ). The type strain is DCY78 = JCM 19896 = KCTC 32174.

128C. hal.per'ni.ae (N.L. gen. n. halperniae of Halpern, named after Malka Halpern, Professor at University of Haifa, Haifa, Israel, whose team isolated E. lactis).

129Basonym: Epilithonimonas lactis Shakéd et al. 2010 (the name Chryseobacterium lactis has already been validly published, hence a new epithet must be chosen to avoid homonyms)

130The description is the same as for Epilithonimonas lactis (Shakéd et al., ). The type strain is DSM 19921 = H1 = LMG 24401.

131C. psy.chro.to'le.rans. (Gr. adj. psychros cold; L. pres. part. tolerans tolerating; N.L. part. adj. psychrotolerans cold-tolerating).

132Basonym: Epilithonimonas psychrotolerans Ge et al. 2015

133The description is the same as for Epilithonimonas psychrotolerans (Ge et al., ) with the following modification. Summed feature 3 should be interpreted as iso-C15:0 2-OH (Montero-Calasanz et al., ). The type strain is CCTCC AB 207182 = NRRL B-51307 = TSBY 57.

134C. te'nax (L. fem. adj. tenax sticky, holding firm).

135Basonym: Epilithonimonas tenax O'Sullivan et al. 2006

136The description is the same as for Epilithonimonas tenax (O'Sullivan et al., ). Lipids were not mentioned in the original description and thus the published corrections for Chryseobacterium (Montero-Calasanz et al., ) do not apply. The type strain is DSM 16811 = EP105 = NCIMB 14026.

137C. xi.xi.so'li (N.L. n. xixi of Xixi, a geographical name; L. gen. n. soli of soil; N.L. fem. gen. n. xixisoli of soil from Xixi).

138Basonym: Epilithonimonas xixisoli Feng et al. 2014

139The description is the same as for Epilithonimonas xixisoli (Feng et al., ). The type strain is CGMCC 1.12802 = NBRC 110387 = S31.

140The description is the same as for the species Flexibacter flexilis, as its known features, while scarce, already differentiate at the genus level.

141On the basis of 16S rRNA gene sequence analysis as well as previously published physiological and morphological data, the type species of the genus represents a separate branch within the order Cytophagales, and must be separated from all other species formerly classified in Flexibacter. The description of the genus must accordingly be restricted.

142The description is as given by Derrien et al. with the following modification. The genomic G+C content of the type strain is 55.8%.

143The description is as given by Nedashkovskaya et al. with the following modification. The genomic G+C content of the type strain is 41.9%.

144The description is as given by Yoon et al. with the following modification. The genomic G+C content of the type strain is 47.8%.

145The description is as given by Rautio et al. with the following modification. The genomic G+C content of the type strain is 53.3%.

146The description is as given by Song et al. with the following modification. The genomic G+C content of the type strain is 57.6%.

147The description is as given by Zhilina et al. with the following restriction. The genomic G+C content of the type strain is 42.7%.

148The description is as given by Downes et al. with the following modification. The genomic G+C content of the type strain is 46.6%.

149The description is as given by Prasad et al. with the following restriction. The genomic G+C content of the type strain is 38.2%.

150The description is as given by Nedashkovskaya et al. with the following modification. The genomic G+C content of the type strain is 36.8%.

151The description is as given by Johnson et al. with the following modification. The genomic G+C content of the type strain is 41.9%.

152The description is as given by Robert et al. with the following modification. The genomic G+C content of the type strain is 42.7%.

153The description is as given by Hayashi et al. with the following modification. The genomic G+C content of the type strain is 45.7%.

154The description is as given by Whitehead et al. with the following modification. The genomic G+C content of the type strain is 35.0%.

155The genome sequence-derived G+C content was reported earlier (Land et al., ) but no taxonomic consequences were drawn.

156The description is as given by Bakir et al. with the following modification. The genomic G+C content of the type strain is 42.0%.

157The description is as given by Holdeman and Moore with the following modification. The genomic G+C content of the type strain is 44.6%.

158The description is as given by Castellani and Chalmers with the following addition. The genomic G+C content of the type strain is 43.1%.

159The description is as given by Nishiyama et al. with the following modification. The genomic G+C content of the type strain is 41.5%.

160The description is as given by Benno et al. with the following modification. The genomic G+C content of the type strain is 44.7%.

161The genome sequence-derived G+C content was reported earlier (Pati et al., ) but no taxonomic consequences were drawn.

162The description is as given by Bakir et al. with the following modification. The genomic G+C content of the type strain is 42.7%.

163The description is as given by Fenner et al. with the following modification. The genomic G+C content of the type strain is 42.7%.

164The description is as given by Eggerth and Gagnon with the following addition. The genomic G+C content of the type strain is 41.9%.

165The description is as given by Ueki et al. with the following restriction. The genomic G+C content of the type strain is 38.0%.

166The description is as given by Castellani and Chalmers with the following addition. The genomic G+C content of the type strain is 42.9%.

167The description is as given by Eggerth and Gagnon with the following addition. The genomic G+C content of the type strain is 46.4%.

168The description is as given by Eggerth and Gagnon with the following addition. The genomic G+C content of the type strain is 42.2%.

169The description is as given by Urios et al. with the following modification. The genomic G+C content of the type strain is 39.8%.

170The description is as given by Morotomi et al. with the following modification. The genomic G+C content of the type strain is 43.9%.

171The description is as given by Brettar et al. with the following modification. The genomic G+C content of the type strain is 36.8%.

172The description is as given by Vandamme et al. with the following restriction. The genomic G+C content of the type strain is 36.1%.

173The description is as given by Schlesner et al. with the following restriction. The genomic G+C content of the type strain is 57.0%.

174The description is as given by Sakamoto et al. with the following restriction. The genomic G+C content of the type strain is 44.8%.

175The description is as given by Sakamoto et al. with the following restriction. The genomic G+C content of the type strain is 42.3%.

176The description is as given by London et al. with the following restriction. The genomic G+C content of the type strain is 40.5%.

177The description is as given by Leadbetter et al. with the following modification. The genomic G+C content of the type strain is 38.4%.

178The description is as given by Bowman with the following restriction. The genomic G+C content of the type strain is 33.8%.

179The genome sequence-derived G+C content was reported earlier (Abt et al., ) but no taxonomic consequences were drawn.

180The description is as given by Johansen et al. with the following modification. The genomic G+C content of the type strain is 32.1%.

181The genome sequence-derived G+C content was reported earlier (Pati et al., ) but no taxonomic consequences were drawn.

182The description is as given by Kämpfer et al. with the following modification. The genomic G+C content of the type strain is 53.0%.

183The description is as given by Sangkhobol and Skerman with the following addition. The genomic G+C content of the type strain is 45.2%.

184The genome sequence-derived G+C content was reported earlier (Glavina Del Rio et al., ) but no taxonomic consequences were drawn.

185The description is as given by Kämpfer et al. with the following modification. The genomic G+C content of the type strain is 44.2%.

186The description is as given by Imhoff with the following restriction. The genomic G+C content of the type strain is 51.3%.

187The description is as given by Imhoff with the following modification. The genomic G+C content of the type strain is 48.4%.

188The description is as given by Gibson et al. with the following restriction. The genomic G+C content of the type strain is 45.0%.

189The description is as given by Kämpfer et al. with the following modification. The genomic G+C content of the type strain is 36.1%.

190The description is as given by Kim et al. with the following modification. The genomic G+C content of the type strain is 33.9%.

191The description is as given by Quan et al. with the following modification. The genomic G+C content of the type strain is 36.6%.

192The description is as given by Young et al. with the following addition. The genomic G+C content of the type strain is 34.8%.

193The description is as given by Kämpfer et al. with the following addition. The genomic G+C content of the type strain is 37.3%.

194The description is as given by Montero-Calasanz et al. with the following restriction. The genomic G+C content of the type strain is 36.8%.

195The description is as given by Loveland-Curtze et al. with the following restriction. The genomic G+C content of the type strain is 34.1%.

196The description is as given by Kämpfer et al. with the following modification. The genomic G+C content of the type strain is 34.9%.

197The description is as given by Kämpfer et al. with the following modification. The genomic G+C content of the type strain is 40.1%.

198The description is as given by Sang et al. with the following modification. The genomic G+C content of the type strain is 38.5%.

199The description is as given by Montero-Calasanz et al. with the following addition. The genomic G+C content of the type strain is 37.3%.

200The description is as given by Pires et al. with the following modification. The genomic G+C content of the type strain is 41.5%.

201The description is as given by Strahan et al. with the following modification. The genomic G+C content of the type strain is 35.2%.

202The description is as given by Weon et al. with the following modification. The genomic G+C content of the type strain is 36.4%.

203The description is as given by Benmalek et al. with the following modification. The genomic G+C content of the type strain is 39.5%.

204The description is as given by Lee et al. with the following modification. The genomic G+C content of the type strain is 41.9%.

205The description is as given by Nedashkovskaya et al. with the following modification. The genomic G+C content of the type strain is 38.3%.

206The description is as given by Raj and Maloy with the following modification. The genomic G+C content of the type strain is 38.1%.

207The description is as given by Shivaji et al. with the following modification. The genomic G+C content of the type strain is 38.8%.

208The description is as given by Reichenbach with the following restriction. The genomic G+C content of the type strain is 37.1%.

209The description is as given by Winogradsky with the following addition. The genomic G+C content of the type strain is 38.8%.

210The description is as given by Dong et al. with the following modification. The genomic G+C content of the type strain is 52.1%.

211The description is as given by Reddy and Garcia-Pichel with the following modification. The genomic G+C content of the type strain is 46.7%.

212The description is as given by Chelius and Triplett with the following modification. The genomic G+C content of the type strain is 51.5%.

213The genome sequence-derived G+C content was reported earlier (Lang et al., ) but no taxonomic consequences were drawn.

214The description is as given by Liu et al. with the following modification. The genomic G+C content of the type strain is 49.0%.

215The description is as given by Hofstad et al. with the following addition. The genomic G+C content of the type strain is 39.6%.

216The description is as given by Lawson et al. with the following modification. The genomic G+C content of the type strain is 37.5%.

217The description is as given by Nedashkovskaya et al. with the following modification. The genomic G+C content of the type strain is 43.8%.

218The description is as given by Nedashkovskaya et al. with the following modification. The genomic G+C content of the type strain is 44.8%.

219The description is as given by Kämpfer et al. with the following addition. The genomic G+C content of the type strain is 35.4%.

220The description is as given by Zhang et al. with the following restriction. The genomic G+C content of the type strain is 32.8%.

221The description is as given by Saha and Chakrabarti with the following modification. The genomic G+C content of the type strain is 35.6%.

222The description is as given by Zhang et al. with the following modification. The genomic G+C content of the type strain is 49.3%.

223The description is as given by Lee et al. with the following modification. The genomic G+C content of the type strain is 47.0%.

224The description is as given by Yoon and Im with the following modification. The genomic G+C content of the type strain is 40.6%.

225The description is as given by Yoon and Im with the following modification. The genomic G+C content of the type strain is 41.3%.

226The description is as given by Yi et al. with the following modification. The genomic G+C content of the type strain is 35.0%.

227The description is as given by Sheu et al. with the following addition. The genomic G+C content of the type strain is 32.2%.

228The description is as given by Dong et al. with the following modification. The genomic G+C content of the type strain is 37.7%.

229The description is as given by Van Trappen et al. (Van Trappen et al., ) with the following modification. The genomic G+C content of the type strain is 34.6%.

230The description is as given by Bernardet et al. with the following modification. The genomic G+C content of the type strain is 32.7%.

231The description is as given by Dong et al. with the following modification. The genomic G+C content of the type strain is 39.6%.

232The description is as given by Dong et al. with the following modification. The genomic G+C content of the type strain is 41.6%.

233The description is as given by Fujii et al. with the following modification. The genomic G+C content of the type strain is 31.2%.

234The description is as given by Larkin et al. with the following modification. The genomic G+C content of the type strain is 37.8%.

235The description is as given by Lewin with the following restriction. The genomic G+C content of the type strain is 42.2%.

236The description is as given by Nedashkovskaya et al. with the following modification. The genomic G+C content of the type strain is 33.5%.

237The description is as given by Macián et al. with the following restriction. The genomic G+C content of the type strain is 36.9%.

238The description is as given by Van Trappen et al. with the following restriction. The genomic G+C content of the type strain is 37.6%.

239The genome sequence-derived G+C content was reported earlier (Riedel et al., ) but no taxonomic consequences were drawn.

240The description is as given by van Veen et al. with the following addition. The genomic G+C content of the type strain is 47.1%.

241The genome sequence-derived G+C content was reported earlier (Daligault et al., ) but no taxonomic consequences were drawn.

242The description is as given by Moore and Moore with the following modification. The genomic G+C content of the type strain is 56.0%.

243The description is as given by Buczolits et al. with the following restriction. The genomic G+C content of the type strain is 62.1%.

244The description is as given by Buczolits et al. with the following addition. The genomic G+C content of the type strain is 56.4%.

245The description is as given Hirsch et al. with the following restriction. The genomic G+C content of the type strain is 56.4%.

246The description is as given by Surendra et al. with the following modification. The genomic G+C content of the type strain is 35.5%.

247The description is as given by Kumar et al. with the following restriction. The genomic G+C content of the type strain is 39.7%.

248The description is as given by Quan et al. with the following modification. The genomic G+C content of the type strain is 33.6%.

249The genome sequence-derived G+C content was reported earlier (Stackebrandt et al., ) but no taxonomic consequences were drawn.

250The description is as given by Weon et al. with the following modification. The genomic G+C content of the type strain is 40.4%.

251The genome sequence-derived G+C content was reported earlier (Abt et al., ) but no taxonomic consequences were drawn.

252The description is as given by Pinhassi et al. with the following modification. The genomic G+C content of the type strain is 39.8%.

253The description is as given by Khan et al. with the following modification. The genomic G+C content of the type strain is 51.6%.

254The description is as given by Bhumika et al. with the following modification. The genomic G+C content of the type strain is 46.7%.

255The description is as given by Lewin with the following modification. The genomic G+C content of the type strain is 40.6%.

256The description is as given by Pankratov et al. with the following modification. The genomic G+C content of the type strain is 42.9%.

257The description is as given by Arun et al. with the following restriction. The genomic G+C content of the type strain is 45.0%.

258The description is as given by Vancanneyt et al. with the following restriction. The genomic G+C content of the type strain is 35.8%.

259The description is as given by Kim et al. with the following modification. The genomic G+C content of the type strain is 48.6%.

260The description is as given by Kwon et al. with the following modification. The genomic G+C content of the type strain is 39.7%.

261The description is as given by Nagai et al. with the following modification. The genomic G+C content of the type strain is 40.5%.

262The description is as given by Hardham et al. with the following addition. The genomic G+C content of the type strain is 43.4%.

263The genome sequence-derived G+C content was reported earlier (Göker et al., ) but no taxonomic consequences were drawn.

264The description is as given by Chin et al. with the following modification. The genomic G+C content of the type strain is 65.3%.

265The description is as given by Zhou et al. with the following modification. The genomic G+C content of the type strain is 40.2%.

266The genome sequence-derived G+C content was reported earlier (Riedel et al., ) but no taxonomic consequences were drawn.

267The description is as given by Sakamoto and Benno with the following modification. The genomic G+C content of the type strain is 45.1%.

268The description is as given by Sakamoto et al. with the following modification. The genomic G+C content of the type strain is 45.2%.

269The description is as given by Sakamoto and Benno with the following modification. The genomic G+C content of the type strain is 45.3%.

270The description is as given by Zhou et al. with the following modification. The genomic G+C content of the type strain is 34.4%.

271The description is as given by Yang et al. with the following modification. The genomic G+C content of the type strain is 40.5%.

272The description is as given by Schlesner and Hirsch with the following addition. The genomic G+C content of the type strain is 57.5%.

273The genome sequence-derived G+C content was reported earlier (Clum et al., ) but no taxonomic consequences were drawn.

274The description is as given by Gosink et al. with the following restriction. The genomic G+C content of the type strain is 32.5%.

275The description is as given by Kim et al. with the following modification. The genomic G+C content of the type strain is 34.5%.

276The description is as given by Nedashkovskaya et al. with the following modification. The genomic G+C content of the type strain is 53.1%.

277The description is as given by Shah and Collins with the following restriction. The genomic G+C content of the type strain is 52.5%.

278The description is as given by Summanen et al. with the following modification. The genomic G+C content of the type strain is 56.3%.

279The description is as given by Willems and Collins with the following modification. The genomic G+C content of the type strain is 51.0%.

280The description is as given by Shah and Collins with the following restriction. The genomic G+C content of the type strain is 47.5%.

281The description is as given by Shah and Collins with the following modification. The genomic G+C content of the type strain is 48.4%.

282The description is as given by Fournier et al. with the following modification. The genomic G+C content of the type strain is 48.6%.

283The description is as given by Shah et al. with the following restriction. The genomic G+C content of the type strain is 45.7%.

284The description is as given by Summanen et al. with the following addition. The genomic G+C content of the type strain is 47.1%.

285The description is as given by Avgustin et al. with the following restriction. The genomic G+C content of the type strain is 41.2%.

286The description is as given by Lawson et al. with the following addition. The genomic G+C content of the type strain is 36.6%.

287The description is as given by Downes et al. with the following modification. The genomic G+C content of the type strain is 53.0%.

288The description is as given by Avgustin et al. with the following restriction. The genomic G+C content of the type strain is 48.7%.

289The description is as given by Avgustin et al. with the following restriction. The genomic G+C content of the type strain is 39.1%.

290The description is as given by Shah and Collins with the following modification. The genomic G+C content of the type strain is 51.0%.

291The description is as given by Hayashi et al. with the following restriction. The genomic G+C content of the type strain is 44.8%.

292The description is as given by Shah and Collins with the following restriction. The genomic G+C content of the type strain is 44.1%.

293The description is as given by Willems and Collins with the following restriction. The genomic G+C content of the type strain is 55.9%.

294The description is as given by Wu et al. with the following restriction. The genomic G+C content of the type strain is 50.1%.

295The description is as given by Downes et al. with the following modification. The genomic G+C content of the type strain is 41.2%.

296The description is as given by Shah and Collins with the following restriction. The genomic G+C content of the type strain is 43.3%.

297The description is as given by Wu et al. with the following restriction. The genomic G+C content of the type strain is 46.6%.

298The description is as given by Downes et al. with the following modification. The genomic G+C content of the type strain is 47.5%.

299The description is as given by Wu et al. with the following addition. The genomic G+C content of the type strain is 41.0%.

300The description is as given by Sakamoto et al. with the following modification. The genomic G+C content of the type strain is 48.3%.

301The genome sequence-derived G+C content was reported earlier (Pati et al., ) but no taxonomic consequences were drawn.

302The description is as given by Alauzet et al. with the following modification. The genomic G+C content of the type strain is 38.4%.

303The description is as given by Shah and Gharbia with the following restriction. The genomic G+C content of the type strain is 42.6%.

304The description is as given by Shah and Collins with the following addition. The genomic G+C content of the type strain is 44.5%.

305The description is as given by Shah and Collins with the following restriction. The genomic G+C content of the type strain is 43.8%.

306The description is as given by Könönen et al. with the following addition. The genomic G+C content of the type strain is 37.4%.

307The description is as given by Ueki et al. with the following modification. The genomic G+C content of the type strain is 37.3%.

308The description is as given by Glazunova et al. with the following addition. The genomic G+C content of the type strain is 42.4%.

309The description is as given by Imhoff with the following addition. The genomic G+C content of the type strain is 50.1%.

310The description is as given by Chen and Dong with the following restriction. The genomic G+C content of the type strain is 43.1%.

311The description is as given by Cao et al. with the following restriction. The genomic G+C content of the type strain is 36.6%.

312The description is as given by Vaz-Moreira et al. with the following modification. The genomic G+C content of the type strain is 38.9%.

313The description is as given by Bowman et al. with the following modification. The genomic G+C content of the type strain is 35.8%.

314The description is as given by Bowman et al. with the following modification. The genomic G+C content of the type strain is 34.5%.

315The description is as given by Donachie et al. with the following restriction. The genomic G+C content of the type strain is 36.5%.

316The description is as given by Bowman et al. with the following restriction. The genomic G+C content of the type strain is 33.4%.

317The description is as given by Schmidt et al. with the following modification. The genomic G+C content of the type strain is 41.8%.

318The description is as given by Collins et al. with the following restriction. The genomic G+C content of the type strain is 57.0%.

319The description is as given by Cho and Giovannoni with the following modification. The genomic G+C content of the type strain is 55.3%.

320The description is as given by Nedashkovskaya et al. with the following modification. The genomic G+C content of the type strain is 40.0%.

321The description is as given by Nedashkovskaya et al. with the following modification. The genomic G+C content of the type strain is 39.3%.

322The description is as given by Lau et al. with the following modification. The genomic G+C content of the type strain is 40.2%.

323The description is as given by Ryu et al. with the following modification. The genomic G+C content of the type strain is 44.3%.

324The description is as given by Larkin and Williams with the following modification. The genomic G+C content of the type strain is 46.6%.

325The genome sequence-derived G+C content was reported earlier (Copeland et al., ) but no taxonomic consequences were drawn.

326The description is as given by Chelius et al. with the following modification. The genomic G+C content of the type strain is 42.1%.

327The description is as given by Ivanova et al. with the following modification. The genomic G+C content of the type strain is 39.8%.

328The description is as given by Chen et al. with the following modification. The genomic G+C content of the type strain is 40.4%.

329The description is as given by An et al. with the following modification. The genomic G+C content of the type strain is 37.4%.

330The description is as given by Yabuuchi et al. with the following modification. The genomic G+C content of the type strain is 39.9%.

331The description is as given by Takeuchi and Yokota with the following modification. The genomic G+C content of the type strain is 43.6%.

332The description is as given by Migula with the following restriction. The genomic G+C content of the type strain is 50.1%.

333The genome sequence-derived G+C content was reported earlier (Lail et al., ) but no taxonomic consequences were drawn.

334The description is as given by Ten et al. with the following modification. The genomic G+C content of the type strain is 48.9%.

335The description is as given by Finster et al. with the following modification. The genomic G+C content of the type strain is 50.4%.

336The description is as given by Stanier with the following addition. The genomic G+C content of the type strain is 36.2%.

337The description is as given by Romanenko et al. with the following modification. The genomic G+C content of the type strain is 32.9%.

338The description is as given by Qiu et al. with the following modification. The genomic G+C content of the type strain is 60.2%.

339The description is as given by Nobre et al. (Tenreiro et al., ) with the following modification. The genomic G+C content of the type strain is 48.8%.

340The description is as given by Zhang et al. with the following modification. The genomic G+C content of the type strain is 35.9%.

341The genome sequence-derived G+C content was reported earlier (Lang et al., ) but no taxonomic consequences were drawn.

342The description is as given by Begum et al. with the following restriction. The genomic G+C content of the type strain is 33.5%.

343HK prepared genomic DNA. TW sequenced the genomes. MH, NI, NK, and SM annotated the genomes. JM and MG phylogenetically analyzed the data. MGL and MG collected the G+C content information. RH, MGL, and MG collected the phenotypic information. RH, JM, MGL, and MG interpreted the results. All authors read and approved the final manuscript.

344This work was performed under the auspices of the US Department of Energy's Office of Science, Biological and Environmental Research Program, and by the University of California, Lawrence Berkeley National Laboratory under contract No. DE-AC02-05CH11231. RH was supported by the German Bundesministerium für Ernährung und Landwirtschaft, grant No. 22016812 for Brian J. Tindall.

345The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.