Reassessing taxonomy and virulence in the Fusobacterium nucleatum group—rebuttal of Fusobacterium animalis clades “Fna C1” and “Fna C2,” genome announcement for Fusobacterium watanabei, and description of Fusobacterium paranimalis sp. nov.

ABSTRACT

There is a considerable interest in the association between Fusobacterium animalis and colorectal cancer (CRC). Recently, it was suggested that this association is valid only for a distinct clade of F. animalis (Fna C2) and that F. animalis strains belonging to another clade (Fna C1) are only associated with the oral cavity. It was further suggested that this made Fna C1 a natural comparator when looking for candidate genes associated with the pathogenicity of Fna C2. Based on such comparisons, three candidate operons enriched in CRC were suggested to explain the strong association of colorectal tumor with F. animalis. In the present paper, we show that major taxonomic errors invalidate the existence of two distinct clades of F. animalis and that Fna C1 is simply a rediscovery and misclassification of Fusobacterium watanabei. We further reassess the phylogenetic structure of the entire Fusobacterium nucleatum group encompassing F. animalis and all known closely related species and confirm the current taxonomy using contemporary phylogenetic principles. We also describe a novel Fusobacterium species more closely related to F. animalis than any other known species, for which we propose the name Fusobacterium paranimalis sp. nov. We further searched for the three proposed candidate virulence operons of F. animalis across the entire F. nucleatum group and show that some or all of these are present in all other species except F. watanabei. We also observe considerable variability of Type V secretion systems (T5SS) by subtype and abundance across the F. nucleatum group.

IMPORTANCE

Considerable resources are being used to study associations between the human microbiota and malignancy. There is a particular interest in the connection between Fusobacterium animalis and colorectal cancer. In this paper, we correct recent taxonomic misconceptions of importance to this research and critically reassess proposed gene candidates for explaining F. animalis pathogenicity. We demonstrate the importance of strict adherence to taxonomic rules when discovering possibly novel phylogenetic groups and emphasize that genome references are still not available for all known bacteria. We reassess the phylogeny of the medically important Fusobacterium nucleatum group including F. animalis using contemporary approaches, provide a genome reference for Fusobacterium watanabei, and describe Fusobacterium paranimalis sp. nov. Our results dispute the concept of using a single closely related comparator phylogenetic group when searching for candidate genes potentially explaining species-specific pathogenicity and show that such comparative approaches can only be meaningful when all relevant related species are included.

INTRODUCTION

There is significant interest around the connection between Fusobacterium animalis and colorectal cancer (CRC). In a recent study comparing presumed F. animalis strains isolated from the human oral cavity and CRC tumors, Zepeda-Rivera et al. (1) concluded that F. animalis can be divided into two distinct clades, called “Fna C1” and “Fna C2,” whereof only “Fna C2,” which includes the F. animalis type strain, is associated with CRC. By categorizing “Fna C1” as a clade of F. animalis not associated with CRC, they frame it as the natural comparative taxonomic group for studying the pathogenic properties of the cancer associated “Fna C2.” The study further implies that the standard species-level bacterial identification in diagnostic microbiology is inadequate for research on the role of the microbiome in cancer and has generated substantial attention.

We found several flaws and taxonomic misconceptions invalidating the tale of two clades within the F. animalis species: (i) when classifying “Fna C1” isolates as a lineage of F. animalis, the authors fail to adhere to basic taxonomic principles, even though these principles are correctly described in the article text; (ii) the authors do not discuss that “Fna C1” is at least as closely related to Fusobacterium vincentii as it is to F. animalis, although this is clear from average nucleotide identity (ANI) calculations provided in their own supplementary material; (iii) they have not included all relevant Fusobacterium species in their analyses, thereby failing to acknowledge that “Fna C1” is simply a rediscovery and misclassification of the valid species Fusobacterium watanabei.

Fusobacterium watanabei, most closely related to F. vincentii and F. animalis, was described by Tomida et al. in 2021 (2) and accepted as a validly published species outside of the International Journal of Systematic and Evolutionary Microbiology later the same year (3). Although Tomida et al. report ANIb and dDDH values for its closest neighboring species within the Fusobacterium nucleatum group as below 92.2% and 49.5%, respectively, no genome was published to accompany the novel species announcement. At present, only partial sequences of the 16S rRNA gene (1,400 bp), together with partial sequences of rpoB (700 bp), gyrB (942 bp), and a zinc protease gene (551 bp), are present in nucleotide databases to support the taxonomy. As the type strain is not represented in genome databases such as NCBI RefSeq or GTDB (4), later genome references for other strains of F. watanabei lack the correct species designation. In GTDB, F. watanabei is currently called Fusobacterium nucleatum_J.

While the failure of Zepeda-Rivera et al. to acknowledge the existence of F. watanabei might be explained by the lack of an available whole-genome reference, their rationale for classifying it as a clade of F. animalis instead of a new species is more difficult to understand. The authors have been challenged on these matters (5, 6), but so far, no correction has been published for their paper (7). We therefore acquired the type strain of F. watanabei from the CCUG strain collection (F. watanabei CCUG 74246T) and sequenced it using a combination of Illumina and Oxford Nanopore sequencing. We then re-analyzed the Fusobacterium genomes from the study by Zepeda-Rivera et al. together with the novel genome for F. watanabei CCUG 74246T using contemporary phylogenetic tools and principles. We also included the genome of a Fusobacterium strain (Vestland19) isolated from blood culture, recently mentioned as a putative new species in a study by Bivand et al. (8). Finally, we reassessed whether the CRC-enriched fusobacterial operons described by Zepeda-Rivera et al. are unique to F. animalis or are also present among other species of the F. nucleatum group.

RESULTS AND DISCUSSION

Re-assessment of the Fusobacterium nucleatum group taxonomy

The “Fusobacterium nucleatum group” is a term encompassing Fusobacterium animalis, Fusobacterium nucleatum, Fusobacterium polymorphum, Fusobacterium vincentii (i.e., the four former subspecies of Fusobacterium nucleatum sensu lato), and other closely related species (9). It is a relevant term since the 16S rRNA gene, commonly used for identification in many settings, including diagnostic microbiology, discriminates poorly between several of these species. To evaluate the F. nucleatum group genetic diversity, all genomes used by Zepeda-Rivera et al. were downloaded. In addition, we sequenced the type strain of F. watanabei, CCUG 74246T, along with a putative novel Fusobacterium species isolated from a positive blood culture at the Dept. of Microbiology, Haukeland University Hospital. Finally, we included available genomes of F. hwasookii (N = 8) and F. simiae (N = 4), two additional species closely related to F. animalis also not included in the data analyses in the Zepeda-Rivera et al. paper. All genomes were included in an all-vs-all ANI calculation using SKANI. A heatmap highlighting the canonical 95% ANI species threshold (Fig. 1A) shows that CCUG 742426T clusters together with the strains erroneously suggested to compose F. animalis Clade 1 (“Fna C1”), confirming that these strains are indeed F. watanabei. A closer look at ANI differences between CCUG 74246T and all other strains in this data set (Fig. 1B) reiterates findings from Zepeda-Rivera et al. that F. watanabei and F. animalis share ~93% ANI between the groups, supporting that they are distinct species as per the 95% ANI definition. Figure 1B also shows that F. vincentii, F. animalis, and F. watanabei strains are all within ~93% ANI from each other, bringing into question why only F. animalis and F. watanabei were selected as comparators in the virulence experiments conducted and why F. vincentii was considered a distinct species and not F. watanabei (aka “Fna C1”). The ANI analyses also support that “Fusobacterium strain Vestland19” represents a novel Fusobacterium species more closely related to F. animalis than any other known species in the F. nucleatum group. For this species, we therefore propose the name Fusobacterium paranimalis sp. nov.

Fig 1.

ANI comparisons between Fusobacterium spp. (A) Heatmap of ANI differences between genomes within the Fusobacterium nucleatum group, as determined by SKANI, with color shading highlighting the canonical 95% ANI species delineation rule. F. watanabei type strain CCUG 74246T is highlighted. (B) Jitter plot showing the ANI of species with multiple strains to CCUG 74246T. (C) Jitter plot showing ANI of species with multiple strains to type strain of F. paranimalis sp. nov.

Beyond ANI—the biological species concept (BSC) and the Fusobacterium nucleatum group

Although ANI analysis is widely used to define bacterial species, this approach is based on arbitrary thresholds. As a consequence, alternative strategies, based on the biological species concept (BSC), have been proposed (10–12). The BSC defines a species as a group of interbreeding individuals that remain reproductively isolated from other groups. In the case of bacteria, gene-flow discontinuities can be used to delineate biological species (10, 13). Because the interruption of gene flow is only partially dependent on the degree of genetic divergence (14), ANI-defined species do not necessarily correspond to species defined by BSC. We, thus, applied two methods based on BSC to delineate the composition of biological species within the F. nucleatum group.

We first used the whole data set of Fusobacterium genomes as an input for populations as clusters of gene transfer (PopCOGenT) (10), which can detect recent gene-flow discontinuities that delineate species. PopCOGenT compares the length distribution of identical regions between pairs of genomes to that expected under a model of clonal evolution, and it provides a measure referred to as “length bias.” PopCOGenT identified 12 genetically isolated ecological units and revealed clusters with excellent congruence to ANI-defined species, with no detectable gene-flow among them (Fig. 2A). PopCOGenT results thus support that F. watanabei and F. animalis are distinct biological species and that F. paranimalis represents a novel species. These data also indicate that, in accordance with ANI analysis, the F. nucleatum genomes belong to two distinct ecological units, as is the case for the two F. pseudoperiodonticum genomes. The genetic heterogeneity of F. pseudoperiodonticum has previously been noticed (6).

Fig 2.

Gene flow between Fusobacterium species. (A) Gene flow network of Fusobacterium genomes. Nodes represent bacterial strains, and edges represent the inferred amount of gene flow between them (expressed in terms of length bias). Edge width and color are proportional to the amount of gene flow between pairs of genomes. Clonal clusters (i.e., strains too closely related) were collapsed in single nodes. (B to G) Homologous recombination between Fusobacterium species. Homoplasy/mutation ratios of different Fusobacterium species are plotted. A single curve indicates the presence of one species, whereas the presence of a second lower curve indicates that two different species are present in the sample. The F. watanabei type strain and a random F. animalis strain were used as the test lineage in the F. animalis + F. watanabei and in the F. watanabei + F. animalis comparisons, respectively.

We next wished to corroborate these results using an approach that relies on the detection of homoplasies, which can persist for long times, and thus measure the effect of both recent and historical gene-flow (13–15). Specifically, we applied ConSpeciFix (13), which measures the ratio of homoplastic (h) to non-homoplastic (m) polymorphisms across core genomes. Based on the h/m parameter, the program estimates whether the tested genomes form a coherent unit, that is, a biological species. We focused on F. animalis and F. watanabei, which showed high h/m values, indicative of widespread gene-flow among core genomes within each species (Fig. 2B and C). The inclusion of F. watanabei CCUG 74246T to the F. animalis population or the addition of one F. animalis genome to the F. watanabei population resulted in sharp decreases in h/m values, confirming that F. animalis and F. watanabei are distinct biological species (Fig. 2D and E). Similarly, inclusion of the F. paranimalis genome to either species caused a sharp decrease in the h/m ratio, confirming that F. paranimalis is genetically isolated from F. animalis and F. watanabei (Fig. 2F and G).

We next aimed to determine whether methods based on core genome alignments were suitable to reconstruct genetic relationships within the Fusobacterium nucleatum complex. We thus used the Genome Taxonomy Database Toolkit (GTDB-Tk) (16) to extract the sequences of 120 core genes present in these fusobacterial genomes. The alignment of the core genes was used to generate a neighbor-net split network, which recapitulated the phylotaxonomic order determined by the analyses above (Fig. S1). Thus, F. animalis and F. watanabei were clearly separated, and F. paranimalis clustered closer to F. animalis, in line with ANI analysis. The split of F. nucleatum into two lineages was also observed in the network. Similar results to those described above were obtained when we used the Genealogies Unbiased By recomBinations In Nucleotide Sequences (Gubbins ) program to construct a phylogenetic tree that accounts for the effect of recombination (Fig. 3) (17), where all mentioned groups clustered independently of each other.

Fig 3.

Distribution of virulence factors among Fusobacterium species. A recombination-aware phylogenetic tree of core genomes calculated by GUBBINS is shown, color-coded by species. Bootstrap support values (expressed in percentage) for relevant nodes are reported. A gene presence/absence matrix for the virulence factors calculated by PPanGGOLIN (https://app.readcube.com/library/) is also reported, along with information about strain isolation sources and the counts of Type V secretion system components estimated by macsyfinder. Color categories are explained in the included legends.

A recent excellent work by Connolly and Kelly (18), including all genomes from Zepeda-Rivera et al. (1) but leaving out F. simiae and F. hwasooki, in phylogenetic and PopCOGenT analyses, found the same species distributions as we did. They did not recognize Fna C1 as F. watanabei, but several observations like this highlight that species designations until this point have been inconsistent and likely inaccurate.

Re-assessment of virulence factor presence in all Fusobacterium nucleatum group members

Zepeda-Rivera et al. framed their “Fna C1” (F. watanabei) as a natural comparator for the CRC-associated “Fna C2” (F. animalis) and do not report whether other closely related species harbor their suggested virulence operons eut, pdu, and the GDAR acid resistance system. Using their rationale, F. paranimalis sp. nov. should represent an even more suitable comparator species in power of being even more closely related to “Fna C2” (F. animalis). However, candidate virulence operons should not be put forward based on a comparison of only two taxons within a larger group of closely related species. Also, the occasional presence of other Fusobacterium spp. (like F. nucleatum, F. vincentii, F. polymorphum, and F. pseudoperiodonticum) in CRC samples (19, 20) could indicate commonalities in virulence gene content among these species.

We therefore searched for the three candidate virulence operons from F. animalis in all other closely related species by running PPanGGOLIN on all strains. Briefly, the software clusters proteins into families of likely orthologs based on the provided identity and coverage scores and keeps track of colocalization of each protein within the genomes included in the analysis. These protein families were then annotated using representative protein sequences of stated candidate virulence factors and presented as a presence/absence matrix along with the Gubbins tree (Fig. 3).

Interestingly, F. polymorphum and F. simiae contain all three operons along with F. animalis. These two species are not associated with CRC, although F. polymorphum appears associated with oral dysplasias (21). These species should, therefore, perform similarly to F. animalis in biochemical tests performed by Zepeda-Rivera et al. Further, the F. hwasooki strains lack only the GDAR system, whereas F. nucleatum sensu stricto, F. pseudoperiodonticum, and F. vincentii lack both GDAR and pdu operons but harbor the eut operon. The single genome of F. paranimalis sp. nov. contains the pdu operon, but lacks eut and GDAR operons.

F. watanabei is the only species lacking all the candidate virulence operons. Our finding that these operons are also harbored by other Fusobacterium species makes us conclude that their presence alone cannot answer why only F. animalis is so strongly associated with CRC. The results together imply that there still are other important pieces to the puzzle, and that although including more species in pan-genome comparisons may complicate the picture, it might also unravel novel genes and operons important to Fusobacterium pathogenicity with higher precision.

The gene(s) encoding secreted FadA (22), appearing in three separate forms (FadA1-3), is present in all strains of our collection and therefore is a likely ubiquitous virulence factor of the F. nucleatum group. FadA is shown to be secreted from the cell by T5aSS/autotransporter Fap2 (22), but, as we and others (21, 23, 24) have found, the T5aSS/autotransporters in Fusobacterium spp. are not easily classified into discrete groups. PPanGGOLIN reported several gene families within the “autotransporter” group, and numerous and genetically different genes are annotated as "autotransporter domain-containing protein," making it difficult to disentangle orthology/paralogy relationships. Single protein searches using, e.g., Diamond, allowed single-reference T5SS proteins (Aim1, RadD, Fap2, fusolisin, and FplA) to map to several loci within each strain, inflating the count, but also limiting the potential count of T5SS due to arbitrary cutoffs to a limited set of reference proteins, which may not represent the full repertoire of T5SS within Fusobacterium spp. For instance, the prototypic Fap2 protein is only found in some strains within the F. animalis population (Fig. 3), which speculatively indicates that also other T5aSS have involvement in FadA transport across the cell wall.

We therefore used Macsyfinder v.2 (25) to identify the number of T5SS in each strain using TXSScan, which identifies secretion systems through the hidden Markov model (HMM) protein profiles. In general, a wide variability in the number of T5SS components was observed, both among and within species (Fig. 3). Moreover, an interesting pattern emerged where T5aSS proteins seem enriched in F. animalis, F. vincentii, F. hwasooki, and F. nucleatum sensu stricto compared to other species. F. watanabei is instead enriched in T5cSS relative to other species, another feature distinguishing it from F. animalis.

A striking difference is seen between T5SS abundance and type distributions, where only one T5dSS (FplA) (26) seems ubiquitous within the F. nucleatum group without large variations. In contrast, T5aSS are more present in numbers, but each “type,” here clustered by moderate cut-offs of 70% sequence identity and 80% coverage, is relatively sparsely present between the strains and species we analyzed.

Speculatively, incompatibility dynamics may be at play, exemplified by the two PPanGGOLIN protein families being most similar to fusolisin (27); these two groups shared ~55% protein sequence identity and could be found in all analyzed species, except F. watanabei. If one of the fusolisin types was present, the other type was seldom concurrently seen. These two proteins were the most similar within the N-terminal and C-terminal ends, and most variation within the substrate-binding sequence was normally located in the middle (28). Determining whether the two fusolisin protein families share the same substrate, like histones (29), falls outside the scope of this article.

In summary, host-microbe interactions between Fusobacterium spp. and human tissues do not simply hinge on one or few virulence factors. An important task is to first identify the important niches and ecological correlates of each species within the group, before detangling what accessory genes do in each Fusobacterium species, in each context. In this regard, Connolly and Kelly (18) again highlight that F. animalis and F. polymorphum seem enriched in cancer tissues and gut lesions in Crohn’s disease by analyzing metagenomes from such sites and that F. vincentii together with F. polymorphum are associated with gingival plaque. Fusobacterium animalis, F. vincentii, and F. polymorphum were also enriched in stool samples from patients with type 2 diabetes. Interestingly, like us, Connolly and Kelly identified a bifurcation in the current F. nucleatum sensu stricto population, of which the smallest group was found more often in stool samples than the larger group.

Genome announcement for Fusobacterium watanabei CCUG 74246T

Fusobacterium watanabei CCUG 74246T was de novo-assembled using Oxford Nanopore Technology (ONT) combined with Illumina NGS data. The GC content was 27%, and the resulting chromosome of 1,981,216 bp included 47 tRNAs, 15 rRNAs, one CRISPR array, and 1816 CDSs when assembled with Bakta. As seen in Fig. 1A and B, CCUG 74246T clusters with other F. watanabei strains, and all other F. watanabei members share ~>98% ANI with CCUG 74246T.

Description of Fusobacterium paranimalis Vestland19T

Fusobacterium paranimalis sp. nov. (Gr. prep. para, beside, near, like; L. gen. n. animalis, of an animal; paranimalis resembling (Fusobacterium) animalis) is a long slender fusiform anaerobic gram-negative rod-shaped bacterium (Fig. S2), closely related to F. animalis. It grows with grayish colonies (0.5–1 mm) after 48 hours of incubation in a strict anaerobe atmosphere at 35 degrees Celsius. The 16S rRNA gene shares 99.8%–100% identity with multiple uncultured references from the human oral cavity in GenBank, including “Fusobacterium sp. oral taxon C10 clone DD027” (accession GU429671) submitted by the Human Oral Microbiome Project (https://homd.org/). It is likely a commensal species from the human oral microbiota.

The type strain is F. paranimalis Vestland19T (NCTC 15132 and DSM 119817). The strain was isolated in a single blood culture bottle from a patient with a suspected but unconfirmed bacterial infection in 2019 who recovered without undergoing antimicrobial treatment. Most likely, the finding represented a transient bacteremia originating from the oral cavity. Using MALDI-ToF MS (Bruker MALDI Biotyper database MBT Compass Library 2023 [12438MSP]) for identification gives a low confidence identification with score 1.8 for F. nucleatum. MALDI-ToF spectra comparisons with closely related type strains are included in File S1 and show that provided an appropriate reference spectrum, F. paranimalis can be unambiguously identified to the species level. The chromosome size is 2, 393, 314 bp and included 48 tRNAs, 14 rRNAs, and 2,230 CDSs when annotated with Bakta. The G + C content is 27.0 mol%. The GenBank accession number for the complete whole-genome sequence is GCA_965278035. The GenBank accession number for the full-length 16S rRNA gene sequence is OZ259485 and for the rpoB gene is OZ259222.

Conclusions

A meaningful search for candidate virulence operons that can explain a perceived species-specific pathogenicity necessitates a correct and comprehensive phylogenetic description coupled with the inclusion of all relevant taxonomic groups in the investigation. Our results confirm the current taxonomy for the F. nucleatum group, including F. watanabei. We decisively reject the division of F. animalis into two distinct clades, as suggested by Zepeda-Rivera et al., a suggestion that was never supported by their data. This further dismantles their argument for using “Fna C1” (F. watanabei) as the natural comparator for “Fna C2” (F. animalis) when searching for candidate genes of potential importance in F. animalis CRC tumor invasion. Our expanded analyses show that the candidate operons set forward by Zepeda-Rivera et al. are not unique to F. animalis and consequently cannot solely explain the enrichment of F. animalis in colorectal tumors. Finally, we confirm the discovery of a novel species within the F. nucleatum group more closely related to F. animalis than any other known member of the cluster for which we propose the name F. paranimalis.

MATERIALS AND METHODS

Sequencing of strains F. watanabei CCUG 74246T and F. paranimalis sp. nov. Vestland19T

The type strain CCUG 74246T was purchased from CCUG in 2024. F. paranimalis sp.nov. Vestland19 was retrieved from a blood culture bottle in the clinical laboratory. After storage in −80°C, the strains were cultivated on acumedia blood agar plates (Neogen). Genomic DNA extraction was performed from bacterial colonies using the ELITe Ingenius DNA extractor (ELITechGroup, Turin, Italy) for Illumina sequencing and TANBead Maelstrom 4800 (Taiwan Advanced Nanotech, Taoyuan, Taiwan), for ONT sequencing. The ONT library protocol was Rapid Sequencing Kit V14 (Oxford Nanopore, UK), and sequencing was performed using a R10.4.1 flongle. Reads were trimmed and basecalled using Dorado v.7.4.14 invoked through MinKNOW. Illumina reads were also produced, using the Nextera XT DNA Library Preparation Kit (Illumina) and the Illumina MiSeq System. The genome was subsequently de novo-assembled using Hybracter v0.10.1 (30) hybrid-single with short-read polishing of the flye assembly. Long-read coverage was downsampled to 100× by Hybracter, and short-read coverage was ~91 for CCUG 74246T and ~224 for Vestland19. The genomes were annotated using Bakta v. 1.9.2 (31).

ANI comparisons

The largest collection of F. watanabei with other Fusobacterium spp. isolates was found in the bioproject associated with the article of Zepeda-Rivera et al. (1), PRJNA549513. Using SKANI v.0.2.2 (32), all strains with >~85% ANI similarity to the type strain CCUG 74246T were included in the ANI pair-to-pair comparison to create Fig. 1A in R using ComplexHeatmap (33). The species within the F. nucleatum group, which fell within ~90% ANI and contained >4 strains, were subsampled to create the dotplot in Fig. 1B and C using ggplot2 (34).

Core gene analyses

All genomes in our data set were used as input data for GTDB-Tk, v2.1.1 (16). This tool allows the identification and extraction of sequence information for a set of 120 single-copy marker genes that are conserved among bacteria using whole-genome sequences as input data. A concatenated alignment based on these core genes was then generated using MAFFT v7.475 (35) with default parameters. A neighbor-net split network was built using SplitsTree v4.16.2 (36) based on HKY85 distances, using parsimony-informative sites, and removing gap sites.

The core gene alignment was also screened for the presence of recombination signals using Gubbins (v.3.0.0) (17). Polymorphic sites outside those recombination regions were then extracted and used to build a recombination-aware phylogeny using fasttree (37) for generating the starting tree and IQ-TREE for the final tree (38). The best substitution model was automatically selected by the ModelFinder tool implemented in IQ-TREE. Robustness of the final tree was assessed using 1,000 bootstrap replicates. All other parameters were set as default.

BSC-based methods

We applied two different methods based on the BSC: PopCOGenT (10) and ConSpeciFix (13).

PopCOGenT defines species boundaries based on recent horizontal gene transfer by searching for stretches of high sequence identity between any pair of genomes. Thus, this tool detects gene flow across the entire genomic space, including both core and accessory genes. PopCOGenT estimates “length bias,” a measure of the observed length distribution of identical regions between pairs of genomes compared to a null expectation of non-recombinogenic evolution (10). Based on this measure, PopCOGenT infers population predictions and generates networks of gene flow, with strains as nodes and the length bias as a measure of their relationship. We ran PopCOGenT using all the 151 Fusobacterium strains in our data set, and we plotted gene flow networks using Cytoscape v3.9.1 (39).

ConSpeciFix instead focuses on a core set of genes shared by a group of genomes. In particular, this tool defines which genes are core and then evaluates the ratio of homoplastic/recombinant polymorphisms (h) to mutations (m) across them. Finally, to determine the species membership, a genome sequence is compared to a set of genomes that are already defined as a single species, and the h/m ratio is calculated for an increasing larger number of genomes. If the tested strain belongs to a different species, h/m ratios will decrease (13).

To understand the relationship between F. animalis, F. watanabei, and the proposed F. paranimalis sp. nov., we ran ConSpeciFix using the “personal comparison” method. We ran two preliminary analyses using F. animalis or F. watanabei separately, and then each species was analyzed adding one strain of the other. Then, we analyzed these two populations by adding the F. paranimalis genome.

Identification of virulence factors

The pangenome of all Fusobacterium strains was calculated by running PPanGGOLIN v2.2.1 (40), using as input protein sequences and GenBank annotation files. Gene family clustering was then computed through the MMseqs2 tool implemented in PPanGGOLIN and with thresholds set to 70% for sequence identity and 80% for sequence coverage, without performing the defragmentation step. All other parameters were set as default. Single-gene searches for T5aSS (Fad2, Aim1, RadD, Fusolisin, and FplA) were done by extracting representative protein sequences from strain ATCC 23726 and searching each genome using diamond blastx (41) with 70% coverage and 70% identity for reference.

The generated pangenome was further used to create a gene presence/absence matrix of virulence genes, selected on the basis of Zepeda-Rivera et al. (1) and by information from the GenBank annotation files. Finally, the presence/absence matrix was plotted along with the Gubbins tree using the ggtree R package.

Identification of all the best candidate Type V secretion system components (i.e., the monomeric autotransporters [T5aSS], the two-partner secretion system [T5bSS], and the trimeric secretion system [T5cSS]) was conducted with MacSyFinder (v2.1.4) (42) by running the TXSScan macsy-models (25) and selecting proteomes as “unordered.”

DATA AVAILABILITY

The accession numbers for all included genomes are provided in Table S1. CCUG 74246T and F. paranimalis sp. nov. are available in BioProject PRJEB85514. CCUG 74246T has accession GCA_965119615.1 and chromosome accession OZ221964; Illumina reads are available as SRA experiment ERX13642389; ONT reads are available as SRA experiment ERX13777046. For Vestland19, the GenBank accession number for the complete whole-genome sequence is GCA_965278035; the GenBank accession number for the full-length 16S rRNA gene sequence is OZ259485; the accession number for the rpoB gene is OZ259222.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/mbio.00941-25.

File S1. mbio.00941-25-s0001.docx.

Generation and comparisons of MALDI-ToF reference spectra.

Fig. S1. mbio.00941-25-s0002.pdf.

Splits tree of Fusobacterium spp.

Fig. S2. mbio.00941-25-s0003.pdf.

Light microscopy of F. nucleatum, F. paranimalis sp. nov., and F. watanabei.

Table S1. mbio.00941-25-s0004.xlsx.

Genomes included in the study.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.