A Genus Definition for <i>Bacteria</i> and <i>Archaea</i> Based on a Standard Genome Relatedness Index. — TEI-tagged view

R. A. BarcopersName, G. M. GarritypersName, J. J. ScottpersName, J. P. AmendpersName, K. H. NealsonpersName, D. EmersonpersName (2020date). “A Genus Definition for <i>Bacteria</i> and <i>Archaea</i> Based on a Standard Genome Relatedness Index..” mBiotitle·j.

Object of study — genusterm·key designation in the taxonomyterm·key of Bacteria and Archaea

Research question — whether AFterm and ANIterm can be used to circumscribe genera objectively

Contribution — a reproducible method of delimiting genera from whole-genome sequences and ANIterm, with the genus inflection pointterm·dom

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

A Genusterm·key Definition for <i>Bacteria</i> and <i>Archaea</i> Based on a Standard Genome Relatedness Index.

Abstract In recent decades, the taxonomyterm·key of Bacteria and Archaea , and therefore genusterm·key designation, has been largely based on the use of a single ribosomal gene, the 16S rRNA geneterm·key, as a taxonomic marker. We propose an approach to delineate genera that excludes the direct use of the 16S rRNA geneterm·key and focuses on a standard genome relatedness index, the average nucleotide identityterm. Our findings are of importance to the microbiology community because the emergent properties of Bacteria and Archaea that are identified in this study will help assign genera with higher taxonomic resolution.

INTRODUCTION At the time of writing, 20,768 bacterial and archaeal species/subspecies and over 3,500 genera with validly published names have been described in the taxonomic literature ( 1 ); however, based on a 16S rRNA guided-phylogenetic approach, there are over 200,000 bacterial and archaeal species and 60,000 genera so far detected in the SILVA database ( 2 , 3 ). Sequence data stored in the Joint Genome Institute (JGI) database, which includes data from other databases, have increased exponentially over the past decade ( 4 ), with >76,000 genomes of bacterial and archaeal isolates, >9,000 metagenome-assembled genomes, and >4,000 single-cell amplified genomes being currently available (database accessed on 10 November 2019). There are >210,000 genome assemblies in GenBank (database accessed on 10 November 2019). Despite the increasing number of genomes, there were only <2,000 genomes of type strainsterm·dom that were publicly available at the start of this study, with 1,003 of these genomes only recently published ( 5 ). Currently, there are <9,000 nonredundant genomes of type strainsterm·dom that are publicly available ( 1 ). The number of genomes of type strainsterm·dom is presently increasing by over 1,000 per year, with recent plans to increase this number substantially ( 6 ); therefore, it is becoming easier to access data that represent taxa of interest. Consequently, this added layer of available information could aid in the formal characterization of microorganisms, of which an essential aspect is the proper assignment of genusterm·key and species (rule 12a of the International Code of Nomenclature of Prokaryotes [ICNP] [ 7 ]).

Historically, DNA-DNA hybridizationterm (DDHterm) has been the “gold standard” for species delineation, with a DDHterm value of ≥70% being recognized as the species boundary between two strains (see reference 8 and other references therein). Stackebrandt and Goebel ( 9 ) conducted a correlation analysis between DDHterm values and 16S rRNA geneterm·key sequence identities, and based on this, they proposed a boundary of 16S rRNA geneterm·key sequence similarity of 97% for species delineation. This value, which is still largely used today for operational taxonomic unit (OTU)-based analysis of microbial communities, has been updated by Stackebrandt and Ebers ( 10 ) to a value between 98.7 and 99.0%, based on a greater amount of available sequence data. Subsequently, as genome sequencing has become common, whole-genome comparisons became possible, leading to the advent of genome relatedness indices such as average nucleotide identityterm (ANIterm) ( 11 ), amino acid identity (AAI) ( 12 ), and digital DDHterm ( 13 ). More recently, Kim et al. ( 14 ) proposed a 16S rRNA geneterm·key sequence similarity threshold value of 98.65% for species delineation, equating to ANIterm values of 95 to 96%, which in turn have been equated to the classical standard species delineation threshold DDHterm value of 70% ( 15 , 16 ). A method that only relies on protein-coding genes (i.e., neither rRNA nor tRNA genes are included in analysis) is the Microbial Species Identifierterm (MiSIterm), which employs both alignment fractionsterm (AFterm) and ANIterm for demarcation of species, recommending threshold AFterm and ANIterm values of 0.6 and 96.5% using complete or nearly complete genomes ( 17 ).

Despite these advancements in resolving species delineation, practical guidelines that incorporate genomic properties to demarcate genera have been lacking even though genusterm·key assignment is key to performing meaningful comparisons regarding the physiology, metabolism, and genomic potential of microbes. Methods to demarcate genera have been proposed that are based on either AAI ( 18 ) or the percentage of conserved proteins (POCP; 19 ). The former method provided a range of AAI values (65 to 72%) that were originally obtained by correlation to a now-outdated 16S rRNA geneterm·key identity threshold for genusterm·key. The POCP method directly relies on the 16S rRNA geneterm·key sequence, which is in some cases insensitive to evolutionary changes in the rest of the genome of a given organism, as revealed by different species sharing >99% identity over the length of this gene. This method also arbitrarily sets a genusterm·key boundary at a POCP value of 50%. Additionally, the generally used arbitrary genusterm·key threshold of 95% 16S rRNA geneterm·key identity has been recently revisited to a lower minimum value of 94.5%, with a median sequence identity of 96.4% and confidence interval of 94.55 to 95.05% ( 3 ). In borderline cases, interpretation of results may be unclear if there are no alternative ways to confirm genusterm·key assignment. This is also the case for microorganisms with multiple highly divergent 16S rRNA genesterm·key ( 20 , 21 ). More recently, a taxonomyterm·key for Bacteria based on highly conserved protein-concatenated phylogeny and the normalization of ranks has been proposed ( 22 ). This normalized taxonomyterm·key uses ∼4% of a genome and directly relies on values of relative evolutionary divergence, a nonstandard index. Here, we propose a novel approach that builds on the MiSIterm method ( 17 ) and provides an objective, mathematically sound, and reproducible method of delimiting genera using whole-genome sequences and ANIterm, a standard genome relatedness index. In addition to identifying genusterm·key boundaries, we introduce the concept of the genus inflection pointterm·dom. We implement this approach by testing a variety of taxonomic groups of Bacteria and Archaea . Furthermore, evidence is presented to support the extensive rearrangement of Bacillusterm·key , Clostridium , Lactobacillusterm·key , and Photorhabdus , among other taxa, using a standard genomic index.

(An earlier version of this article was submitted to an online preprint archive [ 23 ].) RESULTS AND DISCUSSION General assessment.

A total of 3,525 genomes of nonredundant species representing 858 genera in 13 different phyla were used (see Data Set S1 in the supplemental material); 3,331 of these genomes were of type strainsterm·dom, with the rest belonging to species that are >99% identical in 16S rRNA geneterm·key sequence to a type strainterm·dom. Genusterm·key demarcation boundaries were determined for each of the 144 genera that were analyzed in detail, at the taxonomic resolution of order/family ( Data Set S2A ). A subtotal of 3,077 of the species (>87%) were delineated in agreement with current taxonomyterm·key, following AFterm and ANIterm analyses, specifically the genus demarcation boundaryterm·dom. This value increases to 92% if known polyphyletic genera (e.g., Bacillusterm·key , Clostridium , Lactobacillusterm·key , and Pseudomonas ) are excluded. These polyphyletic genera had the greatest number of non-type species at or below the genus demarcation boundaryterm·dom. In terms of disagreement with current taxonomyterm·key, Desulfovibrio had the highest proportion of non-type species at or below the genus demarcation boundaryterm·dom, indicating substantial taxonomic issues with this genusterm·key as well.

List of genomes used in this study. Taxa in bold indicate the taxonomic resolution at which the group was analyzed. Download Data Set S1, XLSX file, 0.5 MB .

(a) Summary of taxonomic groups and genomes included in this study. Taxa in bold indicate the taxonomic resolution at which the group was analyzed. For a detailed species list including accession numbers, see Data Set S1 . (b) AFterm and ANIterm mean and median values associated with genusterm·key demarcation boundaries in Archaea and Bacteria . (c) List of estimated genus inflection pointsterm·dom. For accuracy, this list only includes the genus inflection pointsterm·dom that were consistent with Gompertz and logistic regressions and had, at a minimum, R 2 values approximating 0.90 (quartic regression). Download Data Set S2, XLSX file, 0.1 MB .

The AFterm and ANIterm means of the type and non-type species clusters were significantly different ( P < 0.0001) at the taxonomic level of domain for bacterial ( Fig. 1 ) and archaeal (see Fig. S1 ) genera. For the majority of the genera (94%), the combination of AFterm and ANIterm genomic indices resulted in clustering of type and non-type species. The AFterm and ANIterm means of these clusters were significantly different ( P < 0.05) in 94% and 90% of the cases for AFterm and ANIterm, respectively ( Data Set S2A ). Figure 2 shows the distribution of AFterm and ANIterm values at the taxonomic level of family in Bacteria (see Fig. S2 and S3 for distributions at the taxonomic level of order in Bacteria and of order and family in Archaea , respectively). The 448 identified type strainsterm·dom that would need reclassification directly impact 48 of the 144 genera analyzed in this study. The number of genera that are impacted increases to 62 when the estimated genus inflection pointterm·dom (discussed in detail below) is used.

Distribution of AFterm (a) and ANIterm (b) values of type (T; n = 2,382) and non-type (NT; n = 2,571) species compared to a primary reference in their respective order/family in Bacteria . Significant differences were seen between T and NT species in both cases (nonparametric Wilcoxon test, P < 0.0001). The green line indicates the mean. The blue lines show the standard deviation from the mean.

AFterm (a and b) and ANIterm (c and d) values of type ( n = 1,641) and non-type ( n = 1,554) species compared to primary references within specific taxonomic families in Bacteria . The mean AFterm or ANIterm value is shown by a green line in a–d. The mean of means is denoted by the black line in a–d.

Distribution of AFterm and ANIterm values of type (T; n = 82) and non-type (NT; n = 96) species compared to a primary reference in their respective order/family in Archaea . The green line indicates the mean. The blue lines show the standard deviation from the mean. Download FIG S1, EPS file, 0.9 MB .

(a to d) AFterm (a and b) and ANIterm (c and d) values of type (T; n = 2,382) and non-type (NT; n = 2,571) species compared to primary references within specific taxonomic orders in Bacteria . Download FIG S2, EPS file, 0.7 MB .

The means and medians of all the AFterm and ANIterm values associated with genusterm·key demarcation boundaries as they pertain to Bacteria and Archaea are included in Data Set S2B . Altogether ( n = 144), the AFterm values of genusterm·key demarcation boundaries have a mean of 0.331 (95% confidence interval [CI], 0.308 to 0.354), with a median of 0.345 (25% quartile, 0.206; 75% quartile, 0.444). The ANIterm values of genusterm·key demarcation boundaries have a mean of 73.98% (95% CI, 73.34% to 74.62%), with a median of 73.11% (25% quartile, 70.85%; 75% quartile, 76.56%).

The AFterm values of the estimated genus inflection pointsterm·dom obtained in this study ( n = 28; all from genera in Bacteria ; Data Set S2C ) have a mean of 0.333 (95% CI, 0.305 to 0.362), with a median of 0.349 (25% quartile, 0.281; 75% quartile, 0.371). These values are nearly identical to the AFterm mean (0.330) and median (0.345) of the genusterm·key demarcation boundaries. The ANIterm values of the estimated genus inflection pointsterm·dom obtained in this study have a mean of 73.10% (95% CI, 72.50% to 73.70%), with a median of 73.08% (25% quartile, 72.57%; 75% quartile, 73.93%). These values are also in close proximity to the mean (73.98%) and median (73.11%) ANIterm values of the genusterm·key demarcation boundaries.

We encourage the use of the genus demarcation boundaryterm·dom in conjunction with the estimated genus inflection pointterm·dom that is specific to the taxon that is being analyzed ( Data Set S2A ). The mean and median values that were obtained are only presented here for comparison purposes and in the case that a genus demarcation boundaryterm·dom or estimated genus inflection pointterm·dom cannot be determined (e.g., due to lack of type speciesterm·dom and/or type strainsterm·dom). Several representative cases will be discussed in more detail below.

Bacteria .

To initially test our approach with Gram-positive bacteria, the order Lactobacillales was chosen since it contains a number of well-characterized genera and species of economic importance in human and animal health. A continuum of AFterm and ANIterm values characterizes a typical result when intergeneric and intrageneric species of the order Lactobacillales (e.g., all against all) are compared without distinguishing between type and non-type strains or species ( Fig. 3a ). Because there is no primary taxonomic reference genome to compare to, differentiation between different taxonomic groups is not possible. However, when the type speciesterm·dom of Lactococcus , L. lactis subsp. lactis ATCC 19435 T ( 68 ), is used as the primary reference, a clear distinction can be made between type speciesterm·dom and non-type species ( Fig. 3b ). The type speciesterm·dom of genera in Lactobacillales form a distinct cluster toward lower AFterm and ANIterm values, while the non-type species of Lactococcus cluster toward higher AFterm and ANIterm values ( Fig. 3c and d ). There is no overlap between these two clusters. All species were correctly classified into their corresponding groups by using the genus demarcation boundaryterm·dom.

AFterm and ANIterm pairwise genome comparisons in the order Lactobacillales . (a) Type and non-type strains/species of 15 different genera in the order Lactobacillales were pairwise compared ( n = 6,090 comparisons). (b) Type speciesterm·dom ( n = 37) of genera within the order Lactobacillales (circles) and non-type species ( n = 6) of the genusterm·key Lactococcus (in order Lactobacillales ; triangles) were pairwise compared only to the primary reference, type speciesterm·dom Lactococcus lactis subsp. lactis ATCC 19435 T . The bottom-left quadrant demarcates the boundary between type and non-type species. (c and d) Boxplot diagrams of AFterm (c) and ANIterm (d) values as related to non-type (NT) species of the genusterm·key Lactococcus and type (T) species of genera within the order Lactobacillales (nonparametric Wilcoxon test, P = 0.0003 in both cases). Means are shown in green. Standard deviations are shown in blue.

To test our approach with Gram-negative microorganisms, type speciesterm·dom of genera in the order Alteromonadales and non-type Shewanella spp. were pairwise compared to the type speciesterm·dom Shewanella putrefaciens JCM 20190 T , which is the primary reference ( Fig. 4a ). A distinct clustering is seen between the type speciesterm·dom and non-type species. All species were correctly classified into their corresponding groups by using the genus demarcation boundaryterm·dom.

(a to d) Pairwise genome comparisons to primary references. (a) Type speciesterm·dom of the genusterm·key Shewanella , S. putrefaciens JCM 20190 T . Circles, type speciesterm·dom ( n = 19) of genera within the order Alteromonadales ; triangles, non-type species ( n = 23) of the genusterm·key Shewanella (nonparametric Wilcoxon test, P < 0.0001 with AFterm or ANIterm). (b) Type speciesterm·dom of the genusterm·key Flavobacterium , F. aquatile LMG 4008 T . Circles, type speciesterm·dom ( n = 69) of genera within the family Flavobacteriaceae ; triangles, non-type species ( n = 83) of the genusterm·key Flavobacterium ( P < 0.0001 with AFterm or ANIterm). The inset shows a zoomed-in boxed area. (c) Type speciesterm·dom of the genusterm·key Vibrio , V. cholerae ATCC 14035 T . Circles, type speciesterm·dom ( n = 5) of genera within the family Vibrionaceae ; triangles: non-type Vibrio spp. ( n = 65; P = 0.0003 with AFterm or ANIterm). The inset shows zoomed-in boxed area. (d) Type speciesterm·dom of the genusterm·key Methanosarcina , M. barkeri JCM 10043 T . Circles, type speciesterm·dom ( n = 6) of genera within the family Methanosarcinaceae ; triangles, non-type Methanosarcina spp. ( n = 9; P = 0.0017 with AFterm; P = 0.0018 with ANIterm). In all cases, the bottom-left quadrant demarcates the boundary between type and non-type species. Shown are boxplots of AFterm (center) and ANIterm (right) values as related to non-type (NT) species of genusterm·key analyzed and type speciesterm·dom (T) of genera within the order or family analyzed. Means are shown in green. Standard deviations are shown in blue.

Additional examples show that this trend is also seen with taxonomic families ( Fig. 2 and 4b ). The Flavobacteriaceae were investigated as an example of a diverse family with one of the largest number of validly published genera (currently >160). Despite this, the separate clustering of type speciesterm·dom and non-type species was conserved. When the type speciesterm·dom of the genusterm·key Flavobacterium , Flavobacterium aquatile LMG 4008 T , is used as the primary reference, all type speciesterm·dom (i.e., with available genomes) in the Flavobacteriaceae clustered toward lower AFterm and ANIterm values, while the vast majority of the non-type species of Flavobacterium clustered toward higher AFterm and ANIterm values. The five Flavobacterium spp. that were positioned below the AFterm and ANIterm genus demarcation boundaryterm·dom had 16S rRNA geneterm·key sequence identities of <94% to the primary reference. In a separate example, the type speciesterm·dom Vibrio cholerae ATCC 14035 T was used as the primary reference and compared to non-type species in the genusterm·key Vibrio and other type speciesterm·dom of genera in the family Vibrionaceae ( Fig. 4c ). Clustering is seen between the type speciesterm·dom and non-type species. All species were correctly classified into their corresponding groups by using the genus demarcation boundaryterm·dom, with the exception of Vibrio caribbeanicus ATCC BAA-2122 T , which has a 16S rRNA geneterm·key identity of 92.9% to V. cholerae ATCC 14035 T .

Archaea .

This approach was also tested on an archaeal family, Methanosarcinaceae , in the phylum Euryarchaeota ( Fig. 4d ). The primary reference was the type speciesterm·dom Methanosarcina barkeri JCM 10043 T . The results were consistent with previous examples of bacterial taxa displaying clustering of non-type species within a genusterm·key separately from type speciesterm·dom of other genera within the same family. As in previous examples, all species were classified into their correct higher taxa by using the genus demarcation boundaryterm·dom. Similar results were obtained in the family Haloferacaceae . Additionally, the family Thermoproteaceae in the phylum Crenarchaeota was tested, with Pyrobaculum islandicum DSM 4184 T as a primary reference. A significant distinction between type and non-type species was seen with AFterm ( P = 0.0369) but not with ANIterm ( P = 0.3682) ( Data Set S2A ). Despite this, there was no overlap of the type/non-type clusters when both AFterm and ANIterm were used in combination, and all of the species were classified in agreement with current taxonomyterm·key.

Selected case studies.

In the remainder of the manuscript, we will focus on five selected case studies, each of which has longstanding, historical, taxonomic issues. These cases will be discussed in more detail, with the goal of guiding the reader toward an interpretation of similar scenarios in other taxa. These case studies, along with the examples described above, provide a variety of different scenarios that can be analyzed using our approach.

Genusterm·key assignment in the recently rearranged Hydrogenovibrio - Thioalkalimicrobium - Thiomicrospira cluster was tested as an example of a bacterial group with historical taxonomic issues ( 24 , 25 ) ( Fig. S4A ). Boden et al. ( 26 ) recently provided a detailed evaluation of the characteristics of this cluster, which falls within the radiation of the Piscirickettsiaceae and is proposed to place many of the species into different genera. We tested these newly proposed assignments using the approach described above and the MiSIterm method. When testing genusterm·key assignment to Hydrogenovibrio , the type speciesterm·dom Hydrogenovibrio marinus DSM 11271 T was used as a primary reference ( Fig. 5a ). The single data point that crossed the genus demarcation boundaryterm·dom belonged to Hydrogenovibrio halophilus DSM 15072 T . Aside from this data point, there was no overlap of AFterm and ANIterm values between the type- and non-type species, indicating support for the assignment of H. crunogenus , H. thermophilus , and H. kuenenii to the genusterm·key Hydrogenovibrio . Similarly, the proposed new genusterm·key Thiomicrorhabdus ( 26 ) was also tested ( Fig. S4B ). The results indicate that the new classification is supported by AFterm and ANIterm values, as the type speciesterm·dom of genera in the family Piscirickettsiaceae and Thiomicrorhabdus spp. remain well separated. The reclassification of Thioalkalimicrobium spp. to Thiomicrospira spp. ( 26 ) is also supported by ANIterm and AFterm values ( Fig. S4C ), as these species form a cluster that is well separated from the type speciesterm·dom of genera in the family Piscirickettsiaceae .

Analysis of Hydrogenovibrio , as rearranged by Boden et al. ( 26 ). (a) Pairwise genome comparisons to the primary reference Hydrogenovibrio marinus DSM 11271 T . Non-type species within the genusterm·key Hydrogenovibrio are shown in triangles. The type strainterm·dom H. thermophilus I78 T does not have a sequenced genome; therefore, Hydrogenovibrio sp. strain MA2-6 (>99% pairwise identity in 16S rRNA geneterm·key sequence) was used instead. Similarly, H. crunogenus XCL-2 (>99% pairwise identity in 16S rRNA geneterm·key sequence) ( 69 ) is used instead of the type strainterm·dom H. crunogenus TH-55 T , as TH-55 T does not have a sequenced genome. The type speciesterm·dom of genera within the family Piscirickettsiaceae are shown in circles. The bottom-left quadrant demarcates the boundary between type and non-type species. (b) Maximum likelihood phylogenetic tree based on an alignment of 16S rRNA genesterm·key indicating the phylogenetic positioning of H. halophilus DSM 15072 T . The scale bar indicates 20% sequence divergence. Bootstrap values >70% are shown at the nodes. Accession numbers are shown to the right in parentheses. Tm, Thiomicrospira ; Tmr, Thiomicrorhabdus.

(a to h) Distribution of AFterm and ANIterm values in Archaea at the order (a to d) and family (e to h) levels. (a to d) AFterm (a and b) and ANIterm (c and d) values of non-type species ( n = 96) and type speciesterm·dom ( n = 82) compared to primary references within specific taxonomic orders in Archaea . (e to h) AFterm (e and f) and ANIterm (g and h) values of type (T; n = 69) and non-type (NT; n = 73) species compared to primary references within specific taxonomic families in Archaea . Download FIG S3, EPS file, 2.6 MB .

Pairwise genome comparisons of Thiomicrospira and Thiomicrorhabdus . (a) Pairwise genome comparisons to the type speciesterm·dom Thiomicrospira pelophila DSM 1534 T using the old classification scheme (i.e., prior to rearrangement by Boden et al. [ 26 ]) show clear taxonomic issues and no distinct clustering between type and non-type species. Non-type species within the genusterm·key Thiomicrospira are shown in triangles. Type speciesterm·dom of genera within the family Piscirickettsiaceae are shown in circles. Inset shows a zoomed-in boxed area. (b) Pairwise genome comparisons to Thiomicrorhabdus sp. strain KP2 (>99% pairwise identity to the 16S rRNA geneterm·key sequence of the type speciesterm·dom Thiomicrorhabdus frisia JB-A2 T , which does not have a sequenced genome) using the new reclassification scheme by Boden et al. ( 26 ). Non-type species within the genusterm·key Thiomicrorhabdus are shown in triangles. Type speciesterm·dom of genera within the family Piscirickettsiaceae are shown in circles. The bottom-left quadrant demarcates the boundary between type and non-type species. (c) Pairwise genome comparisons to Thiomicrospira pelophila DSM 1534 T using the new reclassification scheme by Boden et al. ( 26 ). Non-type species within the genusterm·key Thiomicrospira are shown in triangles. Type speciesterm·dom of genera within the family Piscirickettsiaceae are shown in circles, with the exception of Thiomicrorhabdus frisia JB-A2 T , which does not have a sequenced genome; Thiomicrorhabdus sp. KP2 (>99% pairwise identity to the 16S rRNA geneterm·key sequence) was used instead. The bottom-left quadrant demarcates the boundaries between type and non-type species. Download FIG S4, EPS file, 0.3 MB .

Our approach has identified a potential misclassification of H. halophilus (basonym Thiomicrospira halophila ) because it clusters with the type speciesterm·dom of genera in the family Piscirickettsiaceae other than Hydrogenovibrio . Below, we show additional evidence that could support a potential reclassification of H. halophilus . A closer look at the multiple 16S rRNA geneterm·key sequences of H. halophilus revealed that they have 94.0 to 94.2% pairwise identities to the type speciesterm·dom, H. marinus DSM 11271 T . Our results differ from those of Boden et al., who reported 95.7% sequence identity ( 26 ) between H. halophilus and H. marinus . The difference in identity appears to be due to differences in the 16S rRNA geneterm·key sequence between the near-full-length version of Sorokin et al. (1,420 bp, GenBank accession number DQ390450 [ 27 ] and the full-length versions of this gene (1,439 bp each; IMG identifiers [IDs] 2518265101 and 2518265550) originating from a draft genome ( 28 ) and used in this study. The phylogenetic placement of this organism also supports the idea that H. halophilus is distinctly positioned in its own clade as it branches away from the cluster of other Hydrogenovibrio spp. ( Fig. 5b ). One differentiating characteristic of H. halophilus is the DNA G+C content of 56.6%, much higher than the ca. 44% of the other newly classified Hydrogenovibrio spp. Another major differentiating aspect is the higher NaCl optimum/maximum of 1.5 M/3.5 M for H. halophilus versus the lower 0.2 to 0.5/0.6 to 1.2 M for other Hydrogenovibrio spp. Therefore, based on AFterm and ANIterm results and molecular, phylogenetic, and physiological evidence, the assignment of H. halophilus to a new genusterm·key is justifiable. Beyond the taxonomic issue with H. halophilus , we also note the relatively low AFterm and ANIterm values of H. crunogenus and Hydrogenovibrio sp. strain MA2-6 (a strain of H. thermophilus ) in relation to H. marinus . These results in conjunction with the low bootstrap values for branches associated with H. crunogenus and H. thermophilus (<70% in reference 26 and in this study) suggest that there are other taxonomic issues that cannot be resolved at this time in this genusterm·key, in part due to the lack of genomes of type strainsterm·dom (see below for discussion on the genus inflection pointterm·dom).

The Bacillaceae were also investigated as an example of a family that is medically and commercially relevant ( Fig. 6 ) but problematic in terms of taxonomyterm·key. Generally, there were significant differences between non-type Bacillusterm·key species and type speciesterm·dom of genera in the family Bacillaceae when B. subtilis 6051 T ( 29 ) was used as a primary reference ( Data Set S2A ). However, approximately half (59 out of 117) of the non-type Bacillusterm·key spp. clustered with type speciesterm·dom, which strongly suggests that the genusterm·key Bacillusterm·key is in need of taxonomic revision. Of note, the human pathogens Bacillusterm·key anthracis and Bacillusterm·key cereus cluster with type speciesterm·dom of genera in Bacillaceae , indicating that they are at least as genomically different from B. subtilis as are other type speciesterm·dom in this family, warranting a taxonomic rearrangement. As shown above with H. halophilus , a similar argument could be made for the taxonomic rearrangement of each one of these species (e.g., <94.5% identity in 16S rRNA geneterm·key sequence to the type speciesterm·dom B. subtilis 6051). Compared to our analyses of other taxonomic groups, these results are atypical in the sense that the non-type species cluster considerably overlaps the type speciesterm·dom cluster. We highlight the fact that a large number of these specimens were designated non-type species of Bacillusterm·key prior to the advent of the universal use of 16S rRNA geneterm·key as a taxonomic marker in the 1990s (i.e., >20% of all type strainsterm·dom of non-type Bacillusterm·key spp. with available genomes in our data set). For example, B. anthracis was first described in 1872 ( 30 ) and appears on the Approved Lists of Bacterial Names ( 31 ). When the primary reference is changed to Anoxybacillus pushchinoensis K1 T , a type speciesterm·dom of a genusterm·key established in 2000 ( 32 ), using the taxonomic framework of the 16S rRNA geneterm·key, the clustering of type speciesterm·dom of genera in Bacillaceae and non-type species of Anoxybacillus is almost without overlap, with the genus demarcation boundaryterm·dom delineating 97% of the species in agreement with current taxonomyterm·key ( Data Set S2A ). Another example is Lysinibacillus , which was proposed and validly published in 2007 ( 33 ), showing distinct clustering between type speciesterm·dom and non-type species in full agreement with current taxonomyterm·key.

Pairwise genome comparisons to the type speciesterm·dom of the genusterm·key Bacillusterm·key , B. subtilis ATCC 6051 T . Circles, type speciesterm·dom ( n = 30) of genera within the family Bacillaceae ; triangles, non-type Bacillusterm·key spp. ( n = 117). (a) The regression line is based on quartic function, with the data set including the AFterm and ANIterm value of 1:100. The dark-gray-shaded area indicates the 95% confidence interval of the trendline. The light-gray-shaded area indicates the 95% confidence interval of AFterm and ANIterm values. The long dash line demarcates the genusterm·key boundary (left quadrant). The short dash line (left quadrant) extends from the estimated genus inflection pointterm·dom. (b) Zoomed-in boxed area shown in panel a; est., estimated.

The AFterm and ANIterm plots reveal different rates of change at the DNA level. In the few cases in which there are sufficient samples to analyze and/or the distribution of values is more widespread, the AFterm and ANIterm plots reveal a polynomial shape approximating a quartic function, with R 2 values generally at >0.96. A second derivative of this function results in a quadratic function that can be used to detect an inflection point with precision. Estimates based on several examples in this study indicate two inflection points. The overall mean of the first inflection point (AFterm, 0.333; ANIterm, 73.10%; Data Set S2C ) generally approximates the mean genus demarcation boundaryterm·dom of all cases herein presented (AFterm, 0.331; ANIterm, 73.98%; Data Set S2B ). Estimations of this genus inflection pointterm·dom via logistic and Gompertz functions generally agree with the estimations by the preferred quartic function, and their use is recommended for accuracy. The second inflection point, at 89 to 92% ANIterm, is consistently lower than the current ANIterm species threshold (≥95% ANIterm) and is seen with the quartic function in each of the cases in which the genus inflection pointterm·dom was able to be estimated. Related to this, genetic discontinuity between 80 and 95% ANIterm has been previously reported in various studies that included analysis of metagenomes and genomes of isolates ( 34 – 37 ). Recent results obtained by Delmont et al. ( 38 ) reported delineation of SAR 11 populations at ANIterm values lower than 95%, consistent with the ANIterm values of the second inflection points that were observed in this study. Because the second inflection point potentially deals with species delineation, detailed exploration (e.g., using subspecies type strainsterm·dom) and discussion of this topic are outside the scope of this study. However, it is noted that ANIterm values of 89 to 92% largely correspond to 16S rRNA geneterm·key identity values of ≥98.65% (see Fig. 3 in reference 14 ). Also, it is noted that there is a continuum of diversity at the genusterm·key level (e.g., as opposed to what is seen at the species level with genetic discontinuity) in addition to clustering (see Fig. 2 ), adding support to the idea proposed by Palmer et al. ( 39 ) that genetic continuum and genomic cohesiveness are not mutually exclusive, at least at the genusterm·key level.

The estimated genus inflection pointterm·dom serves as a practical guide for maximum AFterm and ANIterm values of a genus demarcation boundaryterm·dom and for the identification of a transition zone. Ideally, and assuming abundance of genomes of type strainsterm·dom and, consequently, data points, the estimated genus inflection pointterm·dom should closely match the genus demarcation boundaryterm·dom. As an example, the genus inflection pointterm·dom in Bacillusterm·key was estimated via the quartic function with AFterm of 0.430 and ANIterm of 73.57% ( Fig. 6a ). However, the majority of the non-type Bacillusterm·key spp. (99 out of 117 [85%]) fall below this inflection point at lower AFterm and ANIterm values, suggesting a potentially profound underestimation of genusterm·key diversity within the family Bacillaceae . An analysis of the pangenome of Bacillusterm·key using this estimated genus inflection pointterm·dom as a guide shows that only non-type Bacillusterm·key spp. above the genus inflection pointterm·dom cluster closely with the type speciesterm·dom B. subtilis (the primary reference), while B. anthracis and B. cereus , which are below both the estimated genus inflection pointterm·dom and genus demarcation boundaryterm·dom, cluster with type speciesterm·dom of other genera in Bacillaceae ( Fig. 7 ), corroborating, as pointed out above, that these two pathogens warrant reclassification and renaming. However, doing so would require considerable care so as not to raise objections over safety issues and a push for conservation of the current names over newly proposed names (see rule 56a in reference 5 and opinion 60 in reference 7 ). These results were further corroborated by phylogenomic analysis of Bacillaceae ( Fig. 7 ).

Pangenome of Bacillusterm·key (in blue). Type speciesterm·dom of genera in the family Bacillaceae (i.e., other than B. subtilis ATCC 6051 T ) that are relatively close to the genusterm·key demarcation AFterm and ANIterm threshold (see Fig. 6 ) are shown in orange. The “Core Bacillusterm·key ” bin includes all Bacillusterm·key spp. with AFterm and ANIterm values that are either above the estimated genus inflection pointterm·dom or within the 99% confidence interval of the estimate. “Core Bacillusterm·key ” does not include B. gobiensis (i.e., AFterm and ANIterm values below estimated genus inflection pointterm·dom), B. anthracis , or B. cereus (i.e., B. anthracis and B. cereus have AFterm and ANIterm values below the genus demarcation boundaryterm·dom). ANIterm and AFterm values were obtained via pairwise genome comparisons to the primary reference, B. subtilis ATCC 6051 T (in purple). Hierarchical clustering was performed on the presence/absence of gene clusters using Euclidean distance and Ward linkage. SCG, single-copy genes. Singleton GC, singleton gene clusters.

It is important to note that the genus inflection pointterm·dom is estimated based on nonlinear regression and only serves as a guide for future taxonomic designation. This is particularly important in the case of a new taxonomic designation with a genome displaying AFterm and ANIterm values that are right above the effective genus demarcation boundaryterm·dom; is it a novel type speciesterm·dom of a new genusterm·key or a novel non-type species of an existing genusterm·key? Having an estimated genus inflection pointterm·dom helps in making this decision. Whereas the genus demarcation boundaryterm·dom is a hard boundary, the genus inflection pointterm·dom represents a soft boundary that highlights a region where the rate of change is starting to decrease at the genome level. Toward higher ANIterm values, this rate then starts to increase, which would correspond to a species inflection point. Thus far, the genus inflection pointterm·dom could be properly estimated only for a few taxonomic groups due to a lack of genomes of type strainsterm·dom and/or actual isolates. Bacillusterm·key pumilus , Bacillusterm·key safensis , Bacillusterm·key altitudinis , Bacillusterm·key zhangzhouensis , Bacillusterm·key xiamenensis , and Bacillusterm·key australimaris are borderline species positioned within the 99% confidence interval of the estimated genus inflection pointterm·dom. Their 16S rRNA geneterm·key identities in relation to the primary reference range from 96.8% to 97.4%, suggesting that the estimated genus inflection pointterm·dom of Bacillusterm·key corresponds to a value that is much higher than the current genusterm·key threshold based on 16S rRNA geneterm·key identity (94.5%).

The taxonomic placements of species within Bacillusterm·key were also compared to a recent taxonomyterm·key that is based on highly conserved, protein-concatenated phylogeny (Genome Taxonomyterm·key Database [GTDB] [ 22 ]). There are currently >280 Bacillusterm·key spp. in the literature with validly published names, but <50% of these species have available sequenced genomes of their type strainsterm·dom. The MiSIterm method and GTDB taxonomyterm·key (accessed on 7 May 2019) indicate that there are 19 and 26 species within Bacillusterm·key , respectively ( Data Set S3A ). The 19 species placed within Bacillusterm·key by the MiSIterm method represent the type strainsterm·dom that are above the genus inflection pointterm·dom. However, if borderline species (i.e., within the confidence interval of the estimated genus inflection pointterm·dom) are also included, the number of species within Bacillusterm·key increases to 26. The great majority of the GTDB classifications of species within Bacillusterm·key (i.e., “d_Bacteria; p_Firmicutes; c_Bacilli; o_Bacillales; f_Bacillaceae; g_Bacillusterm·key” in GTDB) were supported by the MiSIterm method if borderline species were taken into account, with two exceptions, as follows: (i) B. gobiensis , which is considerably below the estimated genus inflection pointterm·dom and outside the confidence interval, and (ii) “B. cellulasensis,” which is not a species with a validly published name.

Comparisons of species designations to GTDB taxonomyterm·key. (a) Species composition of Bacillusterm·key based on MiSIterm and GTDB (only considering type strainsterm·dom). (b) Species composition of Clostridium based on MiSIterm and GTDB (only considering type strainsterm·dom, unless noted otherwise). (c) Species composition of Lactobacillusterm·key based on MiSIterm and GTDB (only considering type strainsterm·dom with available genomes). Download Data Set S3, XLSX file, 0.1 MB .

Clostridium is a genusterm·key with historical taxonomic issues. There are currently 165 Clostridium spp. but only 86 available, nonredundant, sequenced genomes of type strainsterm·dom. The MiSIterm method identified 36 of these Clostridium spp. to be at or below the genus demarcation boundaryterm·dom, indicating numerous taxonomic issues with this genusterm·key. In addition to this, there is a lack of available genomes of Clostridium spp. with ANIterm values between 80 and 100%. The genus inflection pointterm·dom of Clostridium is estimated at an AFterm of 0.252 and ANIterm of 72.72%. There are 21 Clostridium spp. at or above this inflection point. A comparison to GTDB taxonomyterm·key (i.e., limited to type strainsterm·dom; n = 19 in GTDB) again reveals a high level of congruency with the MiSIterm method, matching 74% (14/19; Data Set S3B ) of the Clostridium designations. The points of disagreement are limited to Clostridium cavendishii , Clostridium fallax , Clostridium perfringens , Clostridium ventriculi (note that this is a validly published name but illegitimate [ 40 ]), and Clostridium intestinale , with C. intestinale being a borderline case (i.e., within a 99% confidence interval of the estimated genus inflection pointterm·dom). An updated taxonomyterm·key based on phylogenomics has recently proposed to include these five species in Clostridium cluster I ( sensu stricto ), in agreement with our results ( 41 ).

Xenorhabdus ( 42 ) has been the focus of multiple taxonomic evaluations. Xenorhabdus luminescens had been reclassified to the genusterm·key Photorhabdus , as P. luminescens ( 43 ), based in part on DNA-DNA hybridizationterm analysis, even though this method is not necessarily suitable for genusterm·key demarcation. Separate phylogenetic analyses of both Xenorhabdus and Photorhabdus spp. indicated that in most cases, P. luminescens grouped within the Xenorhabdus cluster ( 44 , 45 ). Now that the genome of P. luminescens has been sequenced, it is evident that it contains multiple 16S rRNA genesterm·key ranging in identity from 99 to 100%. Pairwise comparisons of these genes to the 16S rRNA genesterm·key of the type speciesterm·dom Xenorhabdus nematophila result in identities ranging from 94 to 95%, with a borderline status based on the current 16S rRNA geneterm·key sequence identity genusterm·key threshold of 94.5%. If analysis of AFterm and ANIterm is done on Xenorhabdus using X. nematophila ATCC 19061 T ( 70 ) as the reference, P. luminescens sets the genus demarcation boundaryterm·dom ( Fig. 8a ), with AFterm and ANIterm values that are atypically high (AFterm, 0.465; ANIterm, 75.93%) for a type speciesterm·dom of a genusterm·key other than the primary reference. Alternatively, if the same analysis is done on Photorhabdus using P. luminescens DSM 3368 T as the reference, X. nematophila sets the genus demarcation boundaryterm·dom with identical and atypical high AFterm and ANIterm values ( Fig. 8b ). Despite this, the AFterm and ANIterm analysis does not conflict with current classification of these taxa, as non-type species cluster above the genusterm·key boundary in either case, and type speciesterm·dom of genera in Enterobacterales remain separated from this cluster. However, if the genus inflection pointterm·dom is estimated for Xenorhabdus , it is evident that P. luminescens is well above it, suggesting that P. luminescens should be in the same genusterm·key as the reference X. nematophila ( Fig. 8a ). If the inflection point is estimated for Photorhabdus , similar results are obtained, with X. nematophila clearly placed above the estimated inflection point ( Fig. 8b ). At a minimum, the combined results suggest that X. nematophila and P. luminescens belong to the same genusterm·key. Based on priority of publication, P. luminescens should be considered a later homotypic synonym of X. luminescens ( 42 ). It is noteworthy that bioluminescence is usually one of the properties that is highlighted as a differentiating characteristic of Photorhabdus ; however, at least one strain in this genusterm·key, P. luminescens Q-614, is not bioluminescent ( 46 ). Pangenomic analysis of Xenorhabdus spp. further shows that none of the type speciesterm·dom in Enterobacterales cluster with Xenorhabdus , except P. luminescens , suggesting that Photorhabdus is not as genomically different from X. nematophila as are other type speciesterm·dom in Enterobacterales . This is seen despite the close relatedness of Arsenophonus , Morganella , Moellerella , Proteus , and Providencia to Xenorhabdus , which is evident in the phylogenomic analysis using either single-copy genes or ribosomal genes ( Fig. S5 ).

Close relationship between Photorhabdus and Xenorhabdus as revealed by genome relatedness indices. In both panels, the regression line is based on quartic function. The dark-gray-shaded area indicates the 95% confidence interval of the trendline. The light gray shaded area indicates the 95% confidence interval of AFterm and ANIterm values. The data set includes the AFterm and ANIterm value of 1:100. (a) Pairwise genome comparisons to the type speciesterm·dom of the genusterm·key Xenorhabdus , X. nematophila ATCC 19061 T . Circles, type speciesterm·dom ( n = 44) of genera within the order Enterobacterales ; triangles, non-type Xenorhabdus spp. ( n = 13). (b) Pairwise genome comparisons to the type speciesterm·dom of the genusterm·key Photorhabdus , P. luminescens DSM 3368 T . Circles, type speciesterm·dom ( n = 44) of genera within the order Enterobacterales ; triangles, non-type Photorhabdus spp. ( n = 5). The long dash line demarcates the genusterm·key boundary (left quadrant). The short dash line (left quadrant) extends from the estimated genus inflection pointterm·dom.

Genome clustering of type speciesterm·dom of genera in Enterobacterales based on pangenomic and phylogenomic analyses. Top, only the ribosomal genes were included in the phylogenomic analysis. Bottom, 71 single-copy genes were included in the phylogenomic analysis. Download FIG S5, EPS file, 0.2 MB .

Interestingly, the type speciesterm·dom of Morganella , Moellerella , Proteus , and Providencia are in borderline relationships to Xenorhabdus ( Fig. 8 ), with their respective AFterm and ANIterm values in close proximity to the estimated inflection point. We note that pairwise 16S rRNA geneterm·key sequence identities between X. nematophila and Proteus vulgaris range from 94.5 to 95.0%, again a borderline case as far as genusterm·key assignment is concerned. It is noteworthy that the genusterm·key Xenorhabdus was proposed prior to the advent of 16S rRNA geneterm·key as a taxonomic marker. A more in-depth analysis of all these closely related genera and their potential taxonomic placement, especially in relation to Proteus , should be further investigated as more type strainsterm·dom in these taxa are sequenced.

The last case we will analyze is the known polyphyletic Lactobacillusterm·key , a genusterm·key relevant to many subfields of microbiology, including food and public-health microbiology. Lactobacillusterm·key sensu lato has been proposed to have 24 phylogroups ( 47 ). More recently, Lactobacillusterm·key was proposed to contain 10 phylogroups ( 48 ). However, some difference in opinion to this interpretation has been documented ( 49 ), with a recommendation to apply more explicit criteria for the demarcation of a genusterm·key, in addition to the identification of phylogroup exclusive marker genes. Analysis of Lactobacillusterm·key following the approach herein presented, using L. delbrueckii delbrueckii as a primary reference, resulted in the identification of the genus demarcation boundaryterm·dom (AFterm, 0.075; ANIterm, 72.80%) and estimated genus inflection pointterm·dom (AFterm, 0.349; ANIterm, 74.71%; Fig. 9 ). Significant differences were seen between non-type species of Lactobacillusterm·key and type speciesterm·dom of Lactobacillales in terms of AFterm ( P < 0.0001) but not ANIterm ( P = 0.1624), already highlighting profound taxonomic issues in this genusterm·key. This approach identified 73 Lactobacillusterm·key spp. that are below the genus demarcation boundaryterm·dom, which would strongly suggest reclassification of such species to other genera. A comparison to the GTDB taxonomyterm·key ( Data Set S3C ) shows almost (98%) complete agreement on the placement of these species into other genera (63 out of 64 species; 9 species could not be found on GTDB), with the exception of Lactobacillusterm·key floricola . However, the estimated genus inflection pointterm·dom is much higher than the current genus demarcation boundaryterm·dom, hinting at yet additional taxonomic discrepancies. Indeed, most of the Lactobacillusterm·key spp. (186 out of 193 species considered in this study) have AFterm and ANIterm values that fall below the estimated genus inflection pointterm·dom. Besides L. delbrueckii and its subspecies, only one other species is above the estimated genus inflection pointterm·dom, Lactobacillusterm·key equicursoris . Summarizing the results, Lactobacillusterm·key sensu stricto should currently only contain two species, L. delbrueckii and L. equicursoris . GTDB ( 22 ) and Salvetti et al. ( 48 ) currently have 40 (i.e., “d_Bacteria; p_Firmicutes; c_Bacilli; o_Lactobacillales; f_Lactobacillaceae; g_Lactobacillusterm·key”; GTDB accessed on 27 October 2019) and 33 (i.e., “ L. delbrueckii group”) species within Lactobacillusterm·key , respectively. Some of the species in GTDB's “g_Lactobacillusterm·key” ( 22 ) and Salvetti et al.'s “ L. delbrueckii group” ( 48 ) have very low AFterm values of <0.1 (e.g., L. floricola and L. iners ), indicating <10% alignable region of best-bidirectional hits to the genome of L. delbrueckii delbrueckii , the type speciesterm·dom of Lactobacillusterm·key . These results demonstrate that this genusterm·key is currently not being delineated taking into account a consistent aspect of genomic coherence, as more and more species are classified within Lactobacillusterm·key despite being as genetically distant from the primary reference as are other type speciesterm·dom of genera in Lactobacillale s. Phylogenetically, L. delbrueckii and L. equicursoris form a monophyletic group, as seen by different methods in studies by Morita et al. ( 50 ), Zheng et al. ( 47 ), Sun et al. ( 51 ), Salvetti et al. ( 48 ), and Wittouck et al. ( 52 ). In terms of 16S rRNA identity, the only species that has an identity of >94.5% to L. delbrueckii delbrueckii is L. equicursoris (note that subspecies of L. delbrueckii have identities of >99%). All other Lactobacillusterm·key species in GTDB’s “g_Lactobacillusterm·key” ( 22 ) and Salvetti et al.’s “ L. delbrueckii group” ( 48 ) have identities <94% to L. delbrueckii delbrueckii , with values as low as 86.9% (i.e., L. floricola ; based on a comparison of 1,483 bp). These results highlight that there are still many taxonomic issues to resolve within Lactobacillusterm·key .

Pairwise genome comparisons to the type speciesterm·dom of the genusterm·key Lactobacillusterm·key , L. delbrueckii delbrueckii DSM 20074 T . Circles, type speciesterm·dom ( n = 38) of genera within the order Lactobacillales ; triangles, non-type Lactobacillusterm·key spp. ( n = 192). The regression line is based on quartic function, with the data set including the AFterm and ANIterm value of 1:100. The dark-gray-shaded area indicates the 95% confidence interval of the trendline. The light-gray-shaded area indicates the 95% confidence interval of AFterm and ANIterm values. The long dash line demarcates the genusterm·key boundary (left quadrant). The short dash line (left quadrant) extends from the estimated genus inflection pointterm·dom.

Implications.

The approach we have used to demarcate genera is complementary and does not replace the conventional polyphasic method of circumscribing a genusterm·key, which includes thorough analyses of the full-length 16S rRNA geneterm·key, phylogeny, physiology, and metabolism, among other aspects. The AFterm and ANIterm boundary values for genusterm·key demarcation in a given family or order can be refined as the genomes of more type strainsterm·dom and type speciesterm·dom are sequenced and become publicly available. This approach emphasizes the importance of type strainsterm·dom and type speciesterm·dom in the continuous reevaluation of bacterial and archaeal taxa using genomic indices (e.g., ANIterm or AFterm) because these specimens function as reference points for their respective taxonomic groups and remain available in viable form from multiple public repositories, thus making the approach reproducible, replicable, robust, and generalizable ( 53 ). In this respect, our approach is in line with recommendations regarding taxonomyterm·key, including relevant comparisons to type strainsterm·dom and type speciesterm·dom for the characterization of novel Bacteria and Archaea ( 8 ). It is reiterated that any rearrangement or reclassification of taxa should be in alignment with the ICNP ( 7 ). In particular, a potential rearrangement of type speciesterm·dom must be carefully analyzed, as a genusterm·key can only contain one type speciesterm·dom, and rules and principles of nomenclature must be followed in order to properly do that.

Uncultivated Bacteria and Archaea represented by metagenomic assembled genomes (MAGs) and/or single-cell amplified genomes (SAGs) could potentially be used in the implementation of this approach; however, it is noted that type strainsterm·dom and type speciesterm·dom should be used as references for such analyses, if genomes are available (see “Guidelines,” below). The designation of type strainsterm·dom and type speciesterm·dom is governed by the ICNP and requires a thorough characterization of the microorganism and the deposit of the type strainterm·dom in at least two different culture collections in two different countries. Naturally, the proper characterization, and therefore, the type designation, is in most cases impossible for uncultivated Bacteria and Archaea given the current guidelines set by the ICNP. This could change if genomes are subsequently considered type material, as recently proposed ( 54 ) and discussed ( 55 ). In lieu of type strainsterm·dom and type speciesterm·dom, proper and consistent genomic references must be established in order to analyze taxonomic groups without cultivated representatives as well as rigorous standards for sequence quality. Caution is advised in the interpretation of data resulting from low-quality MAGs since they can represent composite genomes of different strains ( 56 ). In theory, SAGs could serve as good taxonomic frames of reference for uncultivated genera; however, genome incompleteness in SAGs is an issue that would need to be addressed.

The MiSIterm method explicitly excludes tRNA and rRNA genes in order to avoid inflation of AFterm or ANIterm values. However, the current taxonomic framework is largely based on the use of the 16S rRNA geneterm·key for classification. Therefore, the approach herein investigated is not completely independent of the 16S rRNA geneterm·key. The fact that in the majority of the groups tested, with the exception of known polyphyletic genera such as Bacillusterm·key (although significant differences between type and non-type species were still detected in these groups), there was strong concordance of the type speciesterm·dom of genera being delineated from non-type species of the primary reference provides evidence to the success of the use of the 16S rRNA geneterm·key within the taxonomic framework, as far as precision is concerned (i.e., not accuracy). The 16S rRNA geneterm·key has served and could continue to serve the scientific community well, especially in maintaining relative consistency in classifications. However, the estimation of the genus inflection pointterm·dom indicates that the 16S rRNA geneterm·key identity minimum threshold value of 94.5% could underestimate genera diversity in some taxa (e.g., Bacillaceae [ Bacillusterm·key ]) and overestimate it in others (e.g., Enterobacterales [ Photorhabdus ]), especially if multiple nonidentical 16S rRNA genesterm·key are present in the genomes. This uncertainty in delineating genera using the 16S rRNA geneterm·key has also been shown by Yarza et al. ( 3 ), considering that the minimum and median identity values for delineation differ by nearly 2%, at 94.5% and 96.4%, respectively. It is clear that different taxonomic groups are characterized by different AFterm and ANIterm threshold values for genusterm·key demarcation. However, even within established taxonomic groups, the effective AFterm and ANIterm threshold values will still be “moving targets,” dependent on the assignment and reassignment of new type speciesterm·dom. Thus, a single, universal, stationary threshold for genusterm·key delineation will not be able to sensitively resolve genusterm·key assignments for all taxa.

The reader is made aware that variations in bidirectional, matched-pair AFterm values (AF1-2 and AF2-1) exist. These variations are usually small but could be relevant in the interpretation of data, especially in borderline cases. Bidirectional, matched-pair ANIterm values (ANI1-2 and ANI2-1) differ by <0.3% in the majority of the cases, which would make the error bars invisible to the naked eye. A representative figure with corresponding standard deviation values for AFterm is included in the supplemental material ( Fig. S6 ).

Pairwise genome comparisons to the type speciesterm·dom of the genusterm·key Flavobacterium , F. aquatile LMG 4008 T ( Fig. 4b ). Circles, type speciesterm·dom ( n = 71) of genera within the family Flavobacteriaceae ; triangles, non-type species ( n = 82) of the genusterm·key Flavobacterium . The bottom-left quadrant demarcates the boundary between type and non-type species. Error bars indicate one standard deviation from the mean. Download FIG S6, EPS file, 1.3 MB .

The use of ANIterm has been previously stated as not applicable for the demarcation of genera ( 14 , 18 , 19 , 57 ). Our results indicate that the MiSIterm method, and therefore AFterm and ANIterm, can be used to visualize natural breakpoints that can be used to circumscribe genera with objectivity, reproducibility, and high resolution if the guidelines presented in this study are followed. Our study identified potentially misclassified species in numerous genera that could not be previously resolved by alternative methods, including DNA-DNA hybridizationterm, 16S rRNA gene-based phylogeny, and phylogeny using concatenated highly conserved proteins. Adaptation of this method to demarcate higher taxonomic ranks has not been tested, as it is beyond the scope of this study. Nonetheless, such an approach would be warranted. Finally, we highlight that the results demonstrate a conserved genomic coherence at the genusterm·key level for numerous different taxa, shedding light on a fundamental emergent property of Bacteria and Archaea .

MATERIALS AND METHODS ANIterm and AFterm.

ANIterm and alignment fractionterm (AFterm) values were obtained by the Microbial Species Identifierterm (MiSIterm) method using ANIcalculator 2014-127, version 1.0 ( https://aniterm.jgi.doe.gov/html/home.php?page=introduction ) ( 17 ), and also as temporarily implemented in the JGI-Integrated Microbial Genomes and Microbiomes (IMG/M) system ( https://img.jgi.doe.gov/ ) via the Pairwise ANIterm tool (accessed 2017 to 2018). ANIterm, as defined by Varghese et al. ( 17 ), is calculated for a pair of genomes by averaging the nucleotide identity of orthologous genes identified as bidirectional best hits (BBHs), which are the genes that show ≥70% sequence identity and ≥70% alignment of the shorter gene. AFterm, as defined by Varghese et al. ( 17 ), is calculated as a fraction of the sum of the lengths of BBH genes divided by the sum of the lengths of all genes in a genome.

Strains.

Unless otherwise noted, only type strainsterm·dom were used in this study. The NamesforLife Database (NamesforLife, LLC, East Lansing, MI [ 58 ]) was primarily used to retrieve nomenclature (current up to 27 October 2019), nomenclatural history, and taxonomic information about validly published type strainsterm·dom and type speciesterm·dom associated with different genera. Complementary to this, equivalent strain numbers assigned by different biological resource centers (e.g., the German Collection of Microorganism and Cell Cultures [DSMZ]) were searched in their respective online catalogues. In addition to what is mentioned above, primary taxonomic literature sources were used to confirm some of these designations. All strain designations were cross-referenced in at least two databases.

Genomes and 16S rRNA geneterm·key sequences.

Publicly available genomes were obtained from IMG and the National Center for Biotechnology Information (NCBI; https://www.ncbi.nlm.nih.gov/ ). Generally, taxonomic orders/families that have genomes of ≥4 type speciesterm·dom of genera and ≥4 non-type species in a given genusterm·key were considered. Genomes that were unanimously flagged by both IMG and the NCBI as low quality were removed from the data set. About 36% of the genomes included in this study were sequenced as part of the JGI-Genomic Encyclopedia of Bacteria and Archaea , a project that focuses on sequencing the genomes of type strainsterm·dom ( 5 , 59 ). Sequences of the 16S rRNA geneterm·key were obtained from either IMG, the NCBI, or EZBioCloud ( 60 ; https://www.ezbiocloud.net/ ). Alignments were separately performed using the SILVA incremental aligner version 1.2.11 ( 61 ; https://www.arb-silva.de/aligner/ ) and ClustalW via the Geneious platform (version R6; Biomatters, Auckland, New Zealand). Genetic distances were calculated in Geneious. PhyML ( 62 ) was used via the Geneious platform to generate the maximum likelihood phylogenetic tree with the following settings: Hasegawa-Kishino-Yano (HKY85) substitution model, 1,000 bootstraps, estimated transition/transversion ratio, estimated proportion of invariable sites, estimated gamma distribution, and branch lengths and substitution rate optimized.

Pangenome analysis.

Pangenome analysis of Bacillusterm·key and other genera in Bacillaceae were processed in anvi’o (version 5.5) ( 63 ) following the workflow for microbial pangenomics ( http://merenlab.org/2016/11/08/pangenomics-v2/ , last accessed 13 May 2019 [ 64 ]). In brief, we generated contig databases for each genome contig file using the command “anvi-gen-contigs-database.” Prodigal ( 65 ) was used to identify open reading frames, and subsequently, each database was populated with HMM profiles by comparison to a collection of single-copy genes using HMMER ( 66 ). Once contig databases were generated for all genomes, we used “anvi-gen-genomes-storage” to generate a master genome storage database to use in the pangenome analysis. We used the NCBI-BLAST option in “anvi-pan-genome” to calculate gene similarity and MCL ( 67 ) for clustering under the following settings: minbit, 0.5; mcl inflation, 2; and minimum occurrence, 2. For phylogenomic analysis of Bacillusterm·key genomes, we selected a subset of 21 genes based on the following criteria: minimum number of genomes in which gene cluster occurs, 35; maximum number of genes from each genome, 1; maximum functional homogeneity index, 0.9; and minimum geometric homogeneity index, 0.99. We then used the anvi’o command “anvi-get-sequences-for-gene-clusters” to concatenate and align target genes from all genomes and “anvi-gen-phylogenomic-tree” to generate the phylogenomic tree from the concatenated FASTA file. The tree was then rerooted using other genera of Bacillaceae as the outgroup.

Statistical analysis.

All bidirectional, matched-pair values of ANIterm (ANI1-2 and ANI2-1) and AFterm (AF1-2 and AF2-1) were reported as single averaged ANIterm and AFterm values in this study. The nonparametric Wilcoxon test was performed separately for each set of ANIterm and AFterm values to determine significant ( P < 0.05) differences between type and non-type species. All statistical analyses were performed using the statistical software JMP Pro, version 14 (SAS Institute, Inc.).

Rationale.

The nomenclatural type or “type” is defined by the International Code of Nomenclature of Prokaryotes (ICNP) as “that element of the taxon with which the name is permanently associated …” (rule 15 [ 7 ]). The type strainterm·dom acts as the single reference for a given species. Each species and subspecies with a validly published name has a type strainterm·dom that is designated at the time the name is proposed. The vast majority (>98%) are represented by one or more viable deposits that are descended from the original type strainterm·dom, are maintained in pure culture, agree closely with its character in the original description, and are available in one or more public culture collections (rules 18a and 30 [ 7 ]). A type speciesterm·dom, which is represented by its type strainterm·dom, acts as the single reference for a given genusterm·key. Type strainsterm·dom and type speciesterm·dom may not necessarily be the most representative members of a species or genusterm·key, respectively. Rather, type strainsterm·dom and type speciesterm·dom represent the first and often only member of the respective taxa, based on the opinion of the individual proposing each name. The rationale used in this study is that sister species in a given genusterm·key should be relatively similar to the type speciesterm·dom of the cognate genusterm·key, indicating high similarity at the genome level. Therefore, when a type speciesterm·dom of a genusterm·key is compared to a non-type species within the same genusterm·key, the AFterm and ANIterm values should be relatively high. In contrast to this, type speciesterm·dom of different sister genera (i.e., in the same taxonomic family or order) should be relatively dissimilar at the genome level. This dissimilarity should be reflected in relatively low AFterm and ANIterm values.

Guidelines.

In addition to the genome(s) under consideration (e.g., of a cultured strain that needs classification), at least two genomes of type speciesterm·dom in the same family or order are needed, as follows: (i) the primary reference which is the phylogenetically closest relative that is a type speciesterm·dom (i.e., the type speciesterm·dom to which all other microbes will be compared), and (ii) all other type speciesterm·dom (≥1) in the same family or order. Furthermore, all available genomes of non-type species that belong to the same genusterm·key of the primary reference are needed. In order for the results to be meaningful, a species must only be represented by the type strainterm·dom unless its genome is not available, in which case a non-type strain of the same species that is ≥99% identical in 16S rRNA geneterm·key sequence(s) (i.e., all copies of 16S rRNA geneterm·key in the genome) may be used as a proxy until a genome becomes available. The accession numbers for these genomes should be reported. The genus demarcation boundaryterm·dom is set by the highest AFterm and ANIterm values of a type speciesterm·dom (i.e., other than the primary reference) in an order/family.

Interpretation.

Every data point in the AFterm and ANIterm plot is calculated in relation to the primary reference, which is defined above as the type strainterm·dom of the type speciesterm·dom of the specific genusterm·key to be analyzed; therefore, the plot represents genomic similarity to the reference genome. If the primary reference is compared against an identical genome (e.g., or a direct comparison against itself), an AFterm value of 1 and ANIterm value of 100 would result. If the primary reference is compared against relatively dissimilar genomes, the AFterm and ANIterm values will reflect proportionally lower values. Thus, the plots that are generated effectively illustrate how distant other data points are from the primary reference genome in terms of AFterm and ANIterm values, or in other words, how distant other data points are from the upper-right corner (AFterm, 1; ANIterm, 100%).

When all comparisons are made in relation to the primary reference, a type speciesterm·dom (i.e., other than the primary reference) should always cluster with other type speciesterm·dom of genera in the same taxonomic family or order, reflecting relatively low AFterm and ANIterm values. A non-type species should always cluster with other non-type species of the same genusterm·key, reflecting relatively high AFterm and ANIterm values. The distinct clustering of type speciesterm·dom and non-type species should be conserved without any overlap. If a non-type species is positioned within the cluster of type speciesterm·dom in the same taxonomic family or order (i.e., at or below the genus demarcation boundaryterm·dom), this indicates that the non-type species could potentially be assigned to a different genusterm·key, as it is as genomically different from the primary reference, as are other type speciesterm·dom in the same taxonomic family or order. If a type speciesterm·dom is positioned within the cluster of non-type species, there is support for potential reclassification of the type speciesterm·dom within the same genusterm·key of the primary reference, given the accumulation of other supporting evidence (e.g., 16S rRNA geneterm·key identity or phylogeny). It is emphasized that any proposed taxonomic rearrangement or reclassification based on AFterm and ANIterm results will necessarily affect nomenclature; therefore, any corresponding changes in nomenclature resulting from such proposal must be in alignment with the ICNP ( 7 ).

Pangenome analysis of Xenorhabdus and Photorhabdus . Download Text S1, DOCX file, 0.1 MB .