Genome Sequence of Fusobacterium nucleatum Subspecies Polymorphum — a Genetically Tractable Fusobacterium

Fusobacterium nucleatum is a prominent member of the oral microbiota and is a common cause of human infection. F. nucleatum includes five subspecies: polymorphum, nucleatum, vincentii, fusiforme, and animalis. F. nucleatum subsp. polymorphum ATCC 10953 has been well characterized phenotypically and, in contrast to previously sequenced strains, is amenable to gene transfer. We sequenced and annotated the 2,429,698 bp genome of F. nucleatum subsp. polymorphum ATCC 10953. Plasmid pFN3 from the strain was also sequenced and analyzed. When compared to the other two available fusobacterial genomes (F. nucleatum subsp. nucleatum, and F. nucleatum subsp. vincentii) 627 open reading frames unique to F. nucleatum subsp. polymorphum ATCC 10953 were identified. A large percentage of these mapped within one of 28 regions or islands containing five or more genes. Seventeen percent of the clustered proteins that demonstrated similarity were most similar to proteins from the clostridia, with others being most similar to proteins from other gram-positive organisms such as Bacillus and Streptococcus. A ten kilobase region homologous to the Salmonella typhimurium propanediol utilization locus was identified, as was a prophage and integrated conjugal plasmid. The genome contains five composite ribozyme/transposons, similar to the CdISt IStrons described in Clostridium difficile. IStrons are not present in the other fusobacterial genomes. These findings indicate that F. nucleatum subsp. polymorphum is proficient at horizontal gene transfer and that exchange with the Firmicutes, particularly the Clostridia, is common.


INTRODUCTION
Fusobacterium nucleatum is a Gram-negative anaerobic species of the phylum Fusobacteria, numerically dominant in dental plaque biofilms, and important in biofilm ecology and human infectious diseases. Dental plaque is a complex and dynamic microbial community that forms as a biofilm on teeth, and harbors more that 400 distinct species in vivo [1]. F. nucleatum is a prominent component quantitatively and is one of the first Gram-negative species to become established in plaque biofilms [2,3]. It is a central species in physical interactions between Gram-positive and Gramnegative species [4] that are likely to be important in biofilm colonization, and contributes to the reducing conditions necessary for the emergence of oxygen-intolerant anaerobes [5]. F. nucleatum is also one of a small number of oral species that is consistently associated with, and increased in number at, sites of periodontitis [1,2,6], one of the most common infections of humans [7]. It is one of the most common oral species isolated from extraoral infections, including blood, brain, chest, lung, liver, joint, abdominal and obstetrical and gynecological infections and abscesses [6,[8][9][10][11][12][13][14]. Further, F. nucleatum is a common anaerobic isolate from intrauterine infections and has been associated with pregnancy complications including the delivery of premature low birth weight infants [13,[15][16][17][18][19]. Thus, F. nucleatum is a significant pathogen in human infections, including several infections with significant societal impact.
There are five recognized subspecies of F. nucleatum: polymorphum, nucleatum, vincentii, fusiforme, and animalis. DNA sequence analysis of the genome of F. nucleatum subspecies nucleatum (ATCC 25586) (FNN) and vincentii (ATCC 49256) (FNV) indicate clear differences in genetic content between the strains. Three hundred forty-six 346 (17%) and 441 (20%) of the ATCC 25586 and ATCC 49256 open reading frames (ORFs) are absent from the other strain, respectively, and rearrangements are evident in the ORFs present in both strains [20,21]. Phenotypic heterogeneity among F. nucleatum strains has led to the concept that it is a ''species complex'' [22,23]. Taxonomic studies confirm that F. nucleatum ssp. polymorphum ATCC 10953 (FNP) represents a phylogenetic branch separate from the previously sequenced F. nucleatum strains. This branch includes significant human pathogens [12,24,25]. Phenotypic investigations of ATCC 10953 have characterized its uptake and metabolism of amino acids, simple sugars and peptides [26][27][28][29], its physical interaction with epithelial, connective tissue and host immune cells [30][31][32], its ability to modulate host immune cell function including the induction of apoptosis [30,[33][34][35], its ability to enhance the survival of strict anaerobes in biofilm and planktonic multispecies cultures [5,36], and its ability act synergistically with other oral pathogens to enhance virulence in animal model systems [37,38]. In addition, ATCC 10953 harbors a single native plasmid, pFN3 [39], and has been shown to be amenable to gene transfer [40].
The distinct taxonomic status, relatively extensive phenotypic analyses, and genetic transformability of F. nucleatum ssp polymorphum ATCC 10953 suggested that genomic analysis would prove valuable to subsequent studies of this species. Thus, we determined, analyzed, and present the genomic DNA sequence of F. nucleatum subspecies polymorphum strain ATCC 10953. Analysis of the FNP genome revealed that 25% of the genes identified are not represented in the previously sequenced fusobacterial genomes, and that horizontal gene transfer (HGT) has contributed to the evolution of this strain.

Genome anatomy
The FNP genome consists of a single circular chromosome containing 2,429,698 bp ( Figure 1, accession number CM000440) and a single circular plasmid of 11,934 bp ( Figure 2, accession number CP111710). The FNP genome is larger than the FNN genome (2,174,499 bp), and the unfinished FNV genome (2,118,259 bp). The GC content of the chromosome is 26.84%, similar to FNN and FNV (Table 1) and the GC content for the plasmid is 24.53%. There are 2,433 predicted ORFs, 42 pseudogenes, 45 tRNA, 15 rRNA, and 11 ncRNA genes in the FNP genome.
Forty-two of the tRNAS lie in one of seven clusters, each containing from two to fourteen tRNAs. Identical clusters, in both content and internal gene order, are found in FNN. However, the relative position of the clusters in the genome is different. In addition, FNN has two additional asparagine tRNAs, which are associated with rRNA gene regions that have not been fully sequenced in ATCC 1095300 so it is possible that these tRNAs are present in FNP. The tRNAs represent all twenty amino acids. In five cases (Ala, Cys, His, Trp, Tyr), there is only a single tRNA. Many of the tRNAs are duplicated (Asn, Gln, Gly, Leu, Pro, Lys, Ser, Thr, Val) or triplicated (Asp, Glu). The clusters include duplicates as well as singleton tRNAs with no clear pattern associated with the distribution of singletons, duplicates and triplicates. Some tRNAs have multiple copies only one of which is unique, e.g. Gly, Lys, Ser, Val. All four Arg tRNAs, the three Met tRNAs (one is identified as the likely initiator) and the two Phe tRNAs have unique sequences.
A 131 bp repeat sequence occurs seventeen times within the genome ( Figure 3). In several cases, the repeat regions have the potential to encode small hypothetical peptides, which we believe were incorrectly annotated as ORFs in FNN and FNV. All of the 131 nt. repeats from FNP were aligned and used to generate a consensus sequence for additional BLASTN searches. Five nearly complete repeats ($90% of the element's length) and eight repeats with gaps were observed. Sequences corresponding to the 39 half of the element were also found in six locations. In most cases, the repeat sequence was found in intergenic regions and not within coding sequence. One complete copy of the repeat and numerous subsequences were identified in FNN and twenty-eight complete copies plus numerous subsequences were identified in FNV. As in FNP, these fell within intergenic regions. Because the long repeats occur within intergenic regions, it is possible that the sequence is involved in gene regulation, though no particular regulatory motifs were found within the sequence.
Many examples of conservation of gene order, operon structure and gene clustering were observed in the genome. Most of the ribosomal protein genes are organized into operons similar to the L11, L10, str, S10, spc and a operons in Escherichia coli; each encodes between two and eleven ribosomal proteins [41]. Several non-r-protein clusters, similar to conserved non-r-protein gene clusters described in E.coli and other bacteria [42] were also identified in the FNP genome. Table 2 shows the conservation of these gene regions between FNP, FN, Clostridium difficile (NC_009089), Bacillus anthracis str. Ames (NC_003997), E.coli MG1655 (NC_000913). The conserved gene clusters were categorized in five groups according to protein function. Clusters 1-11 in group I are primarily RNA and protein constituents of the ribosome. Group II encodes subunits of the F-type two-sector ATPase. Proteins encoded by the group III clusters are involved in RNA synthesis, modification, transcription, and translation. Cluster 17 in group IV, encodes a spermidine/putrescine ATP binding cassette (ABC) superfamily transporter, while group V contains cluster 18, which codes for the molecular chaperones GroEL and GroES. The inner circle (orange) shows the percent GC calculated using a sliding window of 5 kb. The triangles in the next circle show the location and directionality of tRNAs (red) and ncRNAs (blue). The next tract shows the coordinate scale; this is surrounded by the ORFs, on both strands. ORFs are colored by category, as follows: tan, cell processes; purple, cell structure; red; DNA replication and recombination; blue, general metabolism; green, regulation; yellow, transcription; orange, translation, cyan, transport; fuchsia, virulence; and black, unknown. The IStrons are indicated by the fuchsia arrowheads on the outside circle; the intact IStron is indicated with the star. Plasmid (green) and phage locations (cyan) also appear on the outside circle. doi:10.1371/journal.pone.0000659.g001

Genome Comparisons
Slightly more than 62% of the FNP genes (1514) are found in both of the previously sequenced fusobacterial genomes. Thus, nearly 38% of the genome is either wholly unique to FNP or is shared by FNP and only one of the other two genomes. In terms of coding potential, these 919 genes account for the differences between FNP, FNN, and FNV and thus serve to distinguish FNP. Three comparative lists have been generated: ORFs unique to FNP with respect to FNN and FNV; ORFs common to FNP and FNV, but absent in FNN; and ORFs common to FNP and FNN, but absent in FNV (Table S1).
627 FNP ORFs have no ortholog in either FNN or FNV (Table  S1a), including 106 conserved hypothetical proteins, 287 hypothetical proteins and 9 pseudogenes. Thirty-eight ORFs functioning in transport, including transporters of amino acids, oligopeptides, a siderophore, and divalent metals (Hg 2+ , Cu 2+ , Ni 2+ ) are Alignments of fusobacterial relaxase protein domains to mobilization class consensus motifs [66]. Consensus sequence abbreviations: uppercase letters, conserved; lowercase letters, present in 50% of sites; U or u, bulky hydrophobic residues (I, L, V, M, F, Y and W); -, no consensus at this site; Y, putative active site tyrosine residue. Asterisks (*) above residues indicate identity with consensus sequence. Alignments were performed using Clustal W [102] and then adjusted to best fit the consensus. doi:10.1371/journal.pone.0000659.g002 Table 1. General genome statistics for FNP, FNN and FNV. also unique. Seven additional membrane proteins, two phosphotransferases, and two beta-lactamases are also present only in FNP. Twenty-seven ORFs related to transcriptional regulation are unique to FNP, including a LuxS autoinducer ortholog and two sensor histidine kinases. Additionally, FNP encodes several unique proteins related to DNA modification. These include such functions as methylation, histone acetylation, recombination, integration, topoisomerase, and type I restriction and modification. FNP also contains numerous prophage and transpose genes not found in FNN or FNV. Four Tra conjugation genes (FNP_1868-1871) were found only in FNP. These are adjacent to a region encoding two proteins resembling Type IV secretion components, an outer membrane protein, and four hypothetical proteins. Thus, it is plausible that FNP may have obtained a Type IV secretion system via HGT. This region and the prophage sequences are discussed in more detail, below. Ninety-six ORFs (5 are pseudogenes) are shared between FNP and FNV but are missing from FNN (Table S1b), including 20 conserved hypothetical proteins and 36 fusobacterial conserved hypothetical proteins. Citrate lyase, glutamate-ammonia lyase, serine-pyruvate aminotransferase, cholinephosphate cytidylyltransferase, sulfate reductase and malate dehydrogenase enzymes are shared by FNV and FNP. They also each contain two Nacetylmuramoyl-L-alanine amidases not found in FNN. Five regulatory proteins are common to FNP and FNV, including transcriptional regulators of the MarR, MerR, LysR, and Crp families, and the RNA polymerase sigma factor s 54 .
Two hundred thirteen ORFs (5 are pseudogenes) are found in both FNP and FNN but not in FNV (Table S1c). The total may be inflated, however, because the FNV genome is incomplete and some ORFs may have been missed. This set includes 35 conserved hypothetical proteins and 49 fusobacterial conserved hypothetical proteins. There are 30 transport proteins in this group, including iron transporters, a siderophore transporter, amino acid symporters, ion symporters, and a formate/nitrate transporter. Twelve ORFs related to genetic regulation are in this group, including RpiR and TetR family regulators, Fur, an iron-dependent transcriptional regulator and a response regulator and sensor histidine kinase for the ethanolamine utilization pathway. Also common to FNP and FNN are the GroeSL chaperonins, a cold shock protein, a ribosome-related heat shock protein, and a translation inhibitor. Other proteins of interest in this group are a beta-lactamase, bacterioferrin, at least 12 genes encoding the ethanolamine utilization gene family, an autotransporter/adhesin, an O-antigen assembly gene, 5 glycosyltransferases and the gene for a possible immunosuppressive protein FipA [43,44].

Horizontal Gene Transfer
HGT can be detected by several parametric methods based on deviant nucleotide composition [45,46], dinucleotide frequencies [47], codon usage biases [48][49][50] or patterns inferred by Markov chain analysis [51]. Phylogenetic methods determine a gene's unusual similarity or distribution among organisms by comparing phylogenetic trees of different genes from the genome and assessing the significance of any resulting incongruities. Alternative phylogenetic methods exist that do not reconstruct phylogenetic trees like Clarke's phylogenetic discordance test [52] and Lawrence's [53] rank correlation test. Another reliable inference of recent HGT events is the anomalous phylogenetic distribution method wherein a gene is present in one genome but not found in several closely related genomes [54]. This is the approach used examine genes that had no top BLAST hit to either of the two sequenced fusobacterial genomes, FNN and FNV. Based on BLASTP similarity searches, a total of 1235 ORFs, composed of the 621 FNP ORFs and 9 pseudogenes with no top hits to FNN or FNV (Table S1a) and 608 hypothetical or conserved hypothetical proteins, were graphically plotted to identify clusters that could represent regions of HGT ( Figure 4). About 21% of these, or 255 ORFs, mapped within gene clusters. There were 28 specific regions or islands of interest with clusters of 5 or more genes. Top BLASTP hits for each cluster (Table S2) were examined to determine a consensus genus and species.
One hundred forty of the ORFs (out of 255), or 55% were hypothetical proteins with no matches to other bacterial proteins. Of the remaining 115, 20 ORFs or 17% had top hits to the Clostridia. The most common top hits in this class were to Clostridium tetani, Clostridium thermocellum, Clostridium perfringens, and Desulfitobacterium hafniense. Other ORFs had top hits to other Firmicutes including Bacillus, Streptococcus, Listeria and Enterococcus species. Hits to the archaea Methanosarcina mazei, Methanococcoides burtonii, and Methanothermobacter species, as well as to cyanobacteria Nostoc punctiforme, Trichodesmium erythraeum and Synechocystis sp., were also observed.
A 10 kilobase (kb) region of the FNP genome from nt. 27349 to 37954 (FNP_2111-FNP_2124) appears to have arisen via HGT since this region is not found in the other published fusobacterial genomes and since its GC content (30.4%) is higher than that of the remainder of the genome (26.8%). In the Firmicutes, the clostridial %GC ranges from 28.5 to 30.9 so the DNA may have been acquired from this genus. The region includes 14 predicted genes, 5 of which compose Cluster I (Table S2). Twelve of the genes have no orthologs in FNN or FNV. The genes in this region are homologous to the propanediol utilization locus (pdu) of Salmonella enterica serovar typhimurium, which also arose via HGT [55]. Propanediol is a byproduct produced during the fermentation of fucose [55,56], which has been shown to be present in saliva and metabolized by oral bacteria [57]. Although it appears that FNP is missing the fucose catabolism (fuc) operon, some bacteria such as E. coli secrete propanediol [58], so it is possible that FNP can utilize this propanediol pool.
According to Kapatral et al., the genomes of FNN and FNV lack the necessary enzymes for valine, isoleucine, and leucine biosynthesis [20,21]. A region in the FNP genome carries the ilv/leu operon, which is responsible for the biosynthesis of these amino acids. The predicted products encoded by this locus include dihydroxy-acid dehydratase (IlvD, FNP_0059), threonine ammonia-lyase (IlvA, FNP_0060), acetolactate synthase (IlvB and IlvN, FNP_0061 and FNP_0062), 2-isopropyl malate synthase (LeuA, FNP_0063), 3-isopropyl malate synthase (LeuC and LeuD, FNP_0064 and FNP_0065), isopropyl malate dehydrogenase (LeuB, FNP_0067), and ketol-acid reductoisomerase (IlvC, FNP_0069). The cluster of unique genes (Cluster IV) also includes a small hypothetical protein gene between leuD and leuB (FNP_0066). With the exception of ilvA, all of these genes are missing from the genomes of the other two sequenced fusobacteria. Three additional ilv genes are located at non-adjacent loci in the genome including an additional copy of ilvA (FNP_1302), which is not in the other genomes, and two copies of ilvE (branched-chain-amino-acid transaminase), one that is unique to F. nucleatum ATCC 10953 (FNP_1952) and one that is also found in the other two genomes (FNP_1165).
A prophage genome was identified immediately downstream of an arginine tRNA gene between coordinates 2024189 and 2053649 (28.9% GC) (Figures 1 and 5). The ORFs are not found in FNN or FNV (Table S1a). Forty-two open reading frames (FNP_1662-1703) were predicted in this region spanning four clusters of unique genes (XXI-XXIV), including genes encoding integrase, DNA polymerase, antirepressor, helicase, and terminase proteins. Several genes encoding bacteriophage structural components, such as capsid and tail proteins, were also identified in the region, though 20 of the open reading frames encode hypothetical proteins. The predicted tail proteins were most similar to those in a potential prophage genome in C. tetani E88 [59], while the terminase and packaging proteins were most similar to those in the C. perfringens phage phi3626 [60]. High scoring matches to the nonstructural proteins were found in other gram-positive genomes, such as Bacillus halodurans, Streptococcus mitis, and C. thermocellum. Homologs of 10 of the phage proteins were found in FNV, though only one (a helicase, FNP_1671) was found in FNN. An additional block of phage-like genes map between coordinates 1775962 and 1786106 (FNP_1415-1432) ( Figure 1 and Table S2, Cluster XVIII). Only one of the 19 proteins encoded in this region had orthologs in the other fusobacteria and only two of the proteins (a replication protein and integrase) matched to other bacteriophage sequences.
The region between nts. 2174023 and 2218775 in the FNP genome (FNP_1820-1879, Figure 1), containing 59 genes that include Clusters XXV-XXVII (25.4% GC), is predicted to contain a large conjugal plasmid ( Figure 1, Table S1a, Table  S2). Fifty-five unique genes, not found in FNN or FNP, encode a primase/helicase that could function as a replication initiation protein, topoisomerase, integrase, recombinase, a plasmid partitioning protein, and pseudogenes of a mobilization protein and Five composite ribozyme/transposons, similar to the CdISt IStrons described in C. difficile [63] were identified in the genome (Figure 1). The consensus IStron in FNP is 1811 nt. long and contains a 477 nt. intron followed by an open reading frame encoding the transposase-like protein, TlpB. The FNP IStron is 31% identical to CdISt-C34 from C. difficile and contains four conserved RNA sequences that form the catalytic core of group I introns [64]. All five IStrons in FNP are inserted directly downstream of the pentanucleotide TTGAT, which is the conserved site of insertion of the IS8301 family of transposons [65]. The IStrons have an average G+C content of 29%, consistent with that of both fusobacteria and clostridia. In C. difficile, CdISt1 self-splices to remove itself from the mRNA into which it is inserted. As a result, the insertion does not disrupt expression of the gene. In FNP, we predict that only one copy of the IStron (1361160 to 1362978) is a fully functional element with self-splicing and transposition activities, since the other copies have mutations in either the ribozyme or tlpB regions of the element. Three additional sequences with homology to portions of the ribozyme were identified in the FNP genome and ten additional copies of tlpB-like genes occur in the genome. Homologs of this element were not found in any other organism, including the two strains of Fusobacterium that have been previously sequenced, though TlpB sequences are found in a variety of organisms, including cyanobacteria, Bacillus cereus, Enterococcus, Deinococcus and Exiguobacterium. Thus, it appears that a unique exchange between C. difficile and FNP has occurred.
ATCC 10953 harbors a single plasmid, pFN3 [40] (Figure 2a), which is 11,934 bp in length and has a GC content of 24.53%. Eleven pFN3 ORFs were identified: two possible replication protein genes, a possible resolvase/recombinase gene, a DNA relaxation protein gene and seven hypothetical protein genes. The two replication protein genes (FNP_pFN3g01 and FNP_pFNgo5) have predicted protein sequences with 20-22% identity and 27-32% similarity to the putative replication protein of the F. nucleatum native plasmid pFN1 [40]. The sequence upstream of the pFN3 replication protein gene at 1315 has a sequence (1007 to 1136) characterized by clusters of two overlapping 18 bp repeats (repeat 1: TAATAGTACAAATTTCCC; repeat 2: TAGTACAAAT-TTCCCGAT). Several of the repeats are spaced at 22 bp intervals, suggesting that they may represent replication protein binding sites that are characteristic of the replication origin of iteron-regulated plasmids. The resolvase (FNP_pFN3g09) was identified based on the presence of a N-terminal resolvase domain (pfam02796). The DNA relaxation protein (relaxase) (FNP_pFN3g07) has a relaxase domain and contains the conserved consensus motifs defined for relaxase proteins [66] ( Figure 2b). The pFN3 resolvase and relaxase genes both have potential significance for HGT. Resolvases are important in DNA recombination events, including excision and integration of mobile DNA elements. Relaxase proteins mediate the initiation of conjugal transfer of plasmid DNA. Plasmids that encode relaxases, which are not conjugative themselves, may be mobilized with the additional conjugative functions provided in trans. Two other native F. nucleatum plasmids, pFN1 (AF159249) and pPA52 (AF022647), which are 98% identical, also carry relaxase genes [40]. The occurrence of the relaxase genes suggests the possibility that these plasmids were introduced into F. nucleatum by conjugative processes. Consistent with this mechanism of HGT is the finding that plasmids or DNA sequences related to pPA52 have been detected in 18% of F. nucleatum strains examined [67].

Virulence
We identified 132 predicted proteins that may play a role in fusobacterial virulence ( Table 3). Most of these are found in FNN and FNV, though there are a few notable exceptions. As in the two previously sequenced fusobacterial genomes [20,21], we identified a VacJ homolog (FNP_0314). This protein has been shown to play a role in the intracellular spread of Shigella flexneri [68]. Although its mechanism of action has not been examined, VacJ may play a similar role in FNP since recent evidence suggests that F. nucleatum can invade epithelial cells [32], which may allow dissemination throughout the host to cause infections at non-oral sites [69]. Other previously known virulence factors were identified in FNP including the porin FomA (FNP_0972; not in FNV) [70,71], MviN (FNP_1360), which plays a role in virulence in Salmonella typhimurium [72], TraT (FNP_1881) which provides resistance to complement [73], and VacB (FNP_1921), a ribonuclease involved in virulence gene expression in S. flexneri [74,75]. The strain also carries genes for butyrate fermentation (FNP_0790, 0791, 0969, 0970, 0971, 1762 and either 1467 or 2146). The production of butyrate has been associated with mouth odor and gingival inflammation [76]. We also include FipA (described above) as a virulence factor because of its immunosup-     pressive properties [43,77], though it is most similar to an acetyl-CoA transferase of the butyrate fermentation system. The fipA gene is not present in FNV.
The acquisition of iron from the host environment is an important function of most bacterial pathogens [78]. We have identified 26 predicted proteins involved in iron uptake in FNP. Three proteins, HmuV (FNP_2267), HmuU (FNP_2266), and HmuT (FNP_2269), form a heme ABC superfamily ATP binding cassette transporter while two additional proteins (FNP_2270 and FNP_1765) are probable TonB-dependent heme receptors. There are also two additional iron ATP binding cassette transporters (FNP_428-430 and FNP_1451-1454) and a cobalamin/iron ATP binding cassette transporter (FNP_0398-0341). An Nramp family iron transporter (FNP_1660), and an OfeT family oxidase-dependent iron transporter (FNP_0531) were also annotated. Three hemolysin genes were identified (FNP_0006, FNP_0159, and FNP_0999); two of these (FNP_0159 and FNP_0999) have associated TPS family two-partner secretion proteins (FNP_0155/ 0156 and FNP_1246/1247, respectively). This is similar to what is seen in FNV, but different than FNN, which has three such pairs [21]. Several of the iron transporters (FNP_0339, FNP_0426, FNP_0428, FNP_0531, and FNP_0769) are present in FNN but are missing in FNV; thus FNV may have a diminished requirement for iron or may occupy a different niche. As mentioned previously, a homolog of the ferric uptake regulator, Fur (FNP_2353), was identified in the FNP genome, though it is not present in FNV.
Sixteen possible drug transporters were annotated. These included 7 MOP/MATE family multidrug efflux pumps (FNP_0174, FNP_0640, FNP_0890, FNP_1162, FNP_1207, FNP_1299, and FNP_1596), 2 DMT superfamily drug/metabolite transporters (FNP_0388 and FNP_0622) and 2 RND family antiporters (FNP_0507 and FNP_0508). Our annotation did not permit us to predict the substrates of these transporters but it is likely that many of them are antibiotic transporters. With respect to antibiotic resistance, we also annotated 4 genes predicted to encode beta-lactamases. One of these (FNP_0627) is unique to FNP.
We identified all but one (FN0387) of the 14 outer membrane protein genes described by Kapatral, et. al [20] (two, FNP_1046 and FNP_2283, have been re-annotated as AT family autotransporters) and we discovered a gene encoding OmpW (FNP_1248), which is not found in FNN or FNV. Four potential adhesion proteins were identified including a fibronectin-binding protein homolog (FNP_1337), a possible autotransporter adhesion (FNP_1391), and two proteins (FNP_1880 and FNP_1888) containing von Willebrand (vWF) type A domains. In addition to the Type IV secretion system discussed previously, FNP also carries genes that belong to the Type V secretion system. These include ten autotransporter genes (8 class 1, Type Va; 2 class 2, Type Vc) and the Tps secretion genes tpsA and tpsB (Type Vb) ( Table 3). These are a subset of the genes found in FNN and FNV [62].
Twenty-five ORFs predicted to be involved in the biosynthesis of LPS were identified. This is of interest because F. nucleatum has been shown to have endotoxin activity [23,79]. Unlike FNN, however, FNP does not possess the lic operon, which is predicted to attach choline residues to the LPS [20]; only the licC and licD genes, encoding phosphocholine cytidylyltransferase and a phosphotransferase, respectively, are present. In contrast, FNP does contain genes (FNP_1105-1107) that encode N-acylneuraminate cytidylyltransferase, N-acetyl neuraminate synthase, N-acetylneuraminate synthase and a possible lipooligosaccharide sialyltransferase (FNP_1109) that might incorporate sialic acid into LPS, like FNV (note, however, that FNP_1109 is not found in FNN or FNV). This may facilitate evasion from the host immune response.

Signal Transduction
Six potential two-component signal transduction systems were revealed in FNP and three are of particular interest. An OmpRrelated response regulator maps immediately upstream of a possible sensor histidine kinase gene. While the response regulator (FNP_2108) has homologs in both previously sequenced F. nucleatum genomes [20,21] (FN1261, FNV1053), the sensor protein (FNP_2107) is not present in either of these two genomes. Also unique to this region is an ORF (FNP_2109) immediately downstream of the response regulator that is not found in the other sequenced fusobacterial genomes. This protein appears to contain both sensor histidine kinase and response regulator domains and may represent a fusion of two two-component domains. Another system is a possible ethanolamine two-component regulatory system (FNP_0128 and FNP_0129). This system is present in the FNN genome but is not found in the FNV genome, though it may be located in one of the unfinished regions of that genome. Both FNN and FNV have genes encoding the YesM/YesN twocomponent regulatory system. In FNP the response regulator, YesN (FNP_0212) has been interrupted by the insertion of an IStron, The IStron at this locus is predicted to be inactive so yesN should be non-functional. In addition to the six possible twocomponent systems identified in FNP, there is unmatched response regulator and one unmatched sensor histidine kinase.

Communication with other bacteria
F. nucleatum plays an important role in the formation of oral biofilms, or plaque. F. nucleatum is believed to act as a bridging organism between the Gram-positive early colonizers and the Gram-negative late colonizers [23,36,70]. It has been proposed [80] that some bacteria use a compound known as AI-2 (autoinducer-2) for intra-species communication. AI-2 signaling is also involved with biofilm formation [81]. We have identified the protein responsible for AI-2 synthesis, LuxS (FNP_1558). LuxS activity has been previously reported in ATCC 10953 and in three additional F. nucleatum strains [82]. There is no homolog of luxS in either of the two previously annotated F. nucleatum genomes.

Oxidative stress
Oral anaerobes must cope with oxidative stress to survive and contribute to pathogenesis. Established cultures of FNP are aerotolerent, indicating the presence of mechanisms to detoxify oxygen or oxygen radicals [83].  manage oxidative stress and to maintain reduced conditions is thought to facilitate the survival of other anaerobic pathogens. This facet of F. nucleatum ecology may explain why less aerotolerant organisms such as Porphyromonas gingivalis are increased in number in the presence of F. nucleatum [36]. Aeroprotection of bystander organisms may also explain the synergistic increase in virulence of F. nucleatum when combined with Porphyromonas gingivalis, as compared to either species alone [84,85]. Physiological studies demonstrate that in response to oxidative stress, FNP maintains a reduced environment [5], with increases in NADH oxidase and superoxide dismutase activities [83]. Our analysis of the FNP genome revealed an NADH oxidase (FNP_1794), most closely related to treponemal and streptococcal homologs. An ORF encoding a superoxide dismutase was not identified, though, a rubrerythrin protein (FNP_1721), which confers superoxide dismutase-like activity [86] and has homology to a C. perfringens ruberythrin, is present in FNP. Other FNP proteins of potential importance in oxidative stress include a glutathione peroxidase (FNP_2310) [87,88], a thioredoxin (FNP_0146) [89], a glutaredoxin (FNP_1273) [90], and an alkyl hydroperoxide (FNP_2288) [89,91]. Orthologs of these genes are present in both the FNN and FNV genomes.

Conclusion
Analysis of the genome sequence of F. nucleatum subsp. polymorphum revealed that this microorganism has obtained numerous genes from the Firmicute phylum of bacteria. In particular, many of the regions of FNP that were unique to the Fusobacteria encoded proteins with top BLAST hits to the Clostridia. This is perhaps not unexpected as Firmicutes were the largest phylotype isolated in a gingival bacterial diversity study, and the Clostridia represented the largest class within the Firmicutes that were isolated [1]. It appears that FNP was the recipient of DNA from bacteria that are common to the subgingival niche in HGT events. A contrasting interpretation, presented by Mira, et al. [92], is that Fusobacterium represents a genus with a Clostridial metabolic apparatus that has obtained a gram-negative envelope from the Proteobacteria. If this is the case, then it appears that FNN and FNV have lost a complement of genes that are present in FNP. As was revealed in the comparison of FNN to FNV, the fusobacterial genomes are mosaic in structure.

MATERIALS AND METHODS
Bacterial strain, growth conditions, and DNA isolation F. nucleatum subsp. polymorphum ATCC 10953 is a human strain originally isolated from the inflamed gingiva of an adult male [93]. The strain was acquired as a lyophil from the American Type Culture Collection and recovered on Columbia agar with 5% sheep blood under anaerobic growth conditions. Genomic DNA was prepared from an anaerobically cultivated Columbia broth culture, as previously described [71].

Library construction and sequencing
Genomic DNA of F. nucleatum ATCC 10953 was sheared to 2-6 kb in size using a nebulizer (CIS-US, Inc., Bedford, Mass.), purified from an agarose gel, cloned into a pUC18 derivative sequencing vector, and sequenced as previous described [94].

Sequence assembly
PHRED [95,96] was used to determine the sequence and quality of each base of sequencing reads. Atlas genome assembly tools [97] were used to process reads, to remove repetitive reads, and to bin overlapping reads before assembling the contigs with the program PHRAP [95]. An initial assembly of 54,948 reads gave 146 contigs. Determination of linkages among contigs and gap closing was carried out as previously described [94] except that repetitive reads were put back into the contigs by separated bin assemblies. The rRNAs were not completely sequenced and there are 14 short gaps in the sequence.
The coordinate system for the genome was selected to be similar to other sequenced fusobacteria. In this system, the origin of replication is tentatively localized to the region upstream of an ORF with low identity to DnaA. recF and gyrB are found downstream of this putative dnaA gene. dnaE and repB, which are often clustered with dnaA around the origin of replication, are found elsewhere in the FNP genome, as they are in the other two Fusobacterium genomes

Gene identification and annotation
Gene prediction and manual annotation were performed as previously described [94]. For GeneMark [98] gene predictions, Borrelia burgdorferi was used as model. Predicted proteins with unknown functions were classified into three categories: ''hypothetical protein'' for proteins that do not have matches in the GenBank NR database, ''conserved hypothetical protein'' for proteins that have matches to proteins of unknown functions from organisms outside the genus Fusobacterium, and ''fusobacterial conserved hypothetical protein'' for proteins that have matches only to other fusobacterial proteins of unknown functions.
Transfer RNAs were identified using tRNAscan-SE [99]. Noncoding RNA (ncRNA) elements where identified by BLASTing the known ncRNA sequences of FNN, available at the Rfam database (http://www.sanger.ac.uk/Software/Rfam/), against the FNP genome [100]. The FNP regions containing potential ncRNAs were then analyzed by the Rfam database to identify the exact coordinates of each RNA element.

Genome analysis and identification of possible HGT regions
We performed automatic and manual comparison of best hits to the GenBank NR database and BLAST results to the FNN and FNV genomes to find possible genomic islands and recent horizontally transferred genes in the FNP genome. Genes and regions of possible HGT were examined by GC content, cumulative GC profile, and codon usage bias analysis [101] but this did not add significantly to the conclusions. ClustalW was also used to align DNA sequences and to identify repeat sequences in the genome [102].

Analysis of intergenic regions
After completing annotation of all ORFs called by either GeneMark or Glimmer we decided to confirm that the intergenic regions (IGR) contained no ORFs missed by the ORF calling software. Thus, the entire nucleotide sequence of each IGR was compared to Genbank using BLASTX and those results that had an e value of less than 1610 25 were marked for further analysis by annotators. The annotators defined the exact coordinates of each hit and then followed the same process used for the called ORFs to assign the annotation for each region, resulting in the addition of 50 annotation entries. While this process did result in the identification of a few missed ORFs (5-10), the vast majority of IGR annotations were pseudogenes. Of note concerning this analysis is that it allowed us to extend proteins called as small hypothetical proteins into full-length pseudogenes.

Database submission
The FNP ATCC 10953 genome sequence was deposited in GenBank and assigned accession number CM000440 and the plasmid was assigned accession number CP000710.