Complete genome sequence of the termite hindgut bacterium Spirochaeta coccoides type strain (SPN1T), reclassification in the genus Sphaerochaeta as Sphaerochaeta coccoides comb. nov. and emendations of the family Spirochaetaceae and the genus Sphaerochaeta

Spirochaeta coccoides Dröge et al. 2006 is a member of the genus Spirochaeta Ehrenberg 1835, one of the oldest named genera within the Bacteria. S. coccoides is an obligately anaerobic, Gram-negative, non-motile, spherical bacterium that was isolated from the hindgut contents of the termite Neotermes castaneus. The species is of interest because it may play an important role in the digestion of breakdown products from cellulose and hemicellulose in the termite gut. Here we provide a taxonomic re-evaluation for strain SPN1T, and based on physiological and genomic characteristics, we propose its reclassification as a novel species in the genus Sphaerochaeta, a recently published sister group of the Spirochaeta. The 2,227,296 bp long genome of strain SPN1T with its 1,866 protein-coding and 58 RNA genes is a part of the Genomic Encyclopedia of Bacteria and Archaea project.


Introduction
Strain SPN1 T (= DSM 17374 = ATCC BAA-1237) is the type strain of Spirochaeta coccoides and was isolated from the hindgut contents of the lower dry-wood termite Neotermes castaneus [1,2]. The genus Spirochaeta currently consists of 19 validly named species [3]. The genus name was derived from the Latinized Greek words speira, 'a coil' and chaitê, 'hair', yielding the Neo-Latin 'Spirochaeta', the coiled hair [3]. The species epithet was derived from the neo-Greek words coccos, 'a berry' and eidos, meaning 'shape', yielding the Neo-Latin word coccoides, meaning berry-shaped [1]. Based on the nucleotide sequence of the 16S rRNA gene strain SPN1 T was assigned to the genus Spirochaeta, although its coccoid, non-motile cells differ from the morphology of all known validly named spirochetes [1]. Recently, Ritalahti et al. proposed that Spirochaeta sp. Buddy and Spirochaeta sp. Grapes belonged to the novel genus Sphaerochaeta based on their unique morphology and the 16S rRNA sequence similarity to their closest relatives. The two spherical isolates Spirochaeta sp. Buddy and Spirochaeta sp. Grapes were named Sphaerochaeta globosa and Sphaerochaeta pleomorpha, respectively [4]. On the basis of its morphological, physiological and genomic characteristics, S. coccoides is more closely related to Sphaerochaeta than to the remaining Spirochaeta species, and we therefore propose the placement of S. coccoides SPN1 T into the genus Sphaerochaeta. Here we thus present a summary classification and a set of features for S. coccoides SPN1 T , a description of the complete genome sequencing and annotation, and a proposal to reclassify S. coccoides as a member of the genus Sphaerochaeta as Sphaerochaeta coccoides comb. nov.

Classification and features
A representative genomic 16S rRNA sequence of strain SPN1 T was compared using NCBI BLAST [5,6] under default settings (e.g., considering only the high-scoring segment pairs (HSPs) from the best 250 hits) with the most recent release of the Greengenes database [7] and the relative frequencies of taxa and keywords (reduced to their stem [8]) were determined, weighted by BLAST scores. The most frequently occurring genera were Spirochaeta (57.6%), Sphaerochaeta (39.7%) and Cytophaga (2.7%) (22 hits in total). Regarding the six hits to sequences from other members of the genus, the average identity within HSPs was 90.2%, whereas the average coverage by HSPs was 30.9%. Among all other species, the one yielding the highest score was Spirochaeta bajacaliforniensis (AJ698859), which corresponded to an identity of 90.3% and an HSP coverage of 32.6%. (Note that the Greengenes database uses the INSDC (= EMBL/NCBI/DDBJ) annotation, which is not an authoritative source for nomenclature or classification.) The highest-scoring environmental sequence was AY570600 ('biodegraded Canadian oil reservoir clone PL-16B9'), which showed an identity of 91.0% and an HSP coverage of 85.9%. The most frequently occurring keywords within the labels of all environmental samples which yielded hits were 'microbi' (6.5%), 'mat' (4.5%), 'hypersalin' (3.1%), 'termit' (2.8%) and 'hindgut' (2.6%) (228 hits in total). Environmental samples which yielded hits of a higher score than the highest scoring species were not found. The keywords are partially in agreement with the known environmental preferences of S. coccoides SPN1 T , but the results also indicate that the species itself is rarely found in environmental probes. Figure 1 shows the phylogenetic neighborhood of S. coccoides in a 16S rRNA based tree. The sequences of the three 16S rRNA gene copies in the genome differ from each other by up to two nucleotides, and differ by up to two nucleotides from the previously published 16S rRNA sequence (AJ698092).
In contrast to all other validly described spirochete species (except for those currently placed in the novel genus Sphaerochaeta [4]) the cells of S. coccoides SPN1 T are cocci (0.5 to 2.0 µm in diameter) which are surrounded by an outer envelope. In the early growth phase cell aggregates are formed [1]. S. coccoides is a Gram-negative, non-motile and strictly anaerobic bacterium (Table 1). Strain SPN1 T showed no catalase activity [1], although a gene probably coding a catalase (Spico_0266) was identified in the genome. The optimal growth temperature of strain SPN1 T is 30°C, with no growth observed above 40°C or below 15°C [1]. The pH range for growth is 5.5-9.5, with an optimum at pH 7.4 [1]. Maltose is fermented to ethanol, with formate and acetate as the main fermentation products. Glucose, galactose, lactate, pyruvate, amino acids, and polysaccharides are not utilized, but the organism is able to grow with yeast extract as the sole carbon and energy source [1]. A minimum yeast concentration of 0.2% was required for growth [1]. Activities of β-D-glucosidase, α-Dglucosidase, α-D-galactosidase, α-L-arabinosidase, β-D-fucosidase, and β-D-xylosidase are exhibited [1]. These enzymatic activities seemed to be cellbound, as no glycolytic activity was found in the supernatant of the culture [1].

Genome project history
This organism was selected for sequencing on the basis of its phylogenetic position [34], and is part of the Genomic Encyclopedia of Bacteria and Archaea project [35]. The genome project is deposited in the Genomes On Line Database [15] and the complete genome sequence is deposited in GenBank. Sequencing, finishing and annotation were performed by the DOE Joint Genome Institute (JGI). A summary of the project information is shown in Table 2.

Growth conditions and DNA isolation
S. coccoides strain SPN1 T , DSM 17374, was grown anaerobically in DSMZ medium 1204 (Spirochaeta coccoides medium) [36] at 30°C. DNA was isolated from 0.5-1 g of cell paste using MasterPure Gram-positive DNA purification kit (Epicentre MGP04100) following the standard protocol as recommended by the manufacturer with modification st/DL for cell lysis as described in Wu et al. 2009 [35]. DNA is available through the DNA Bank Network [37].

Figure 1.
Phylogenetic tree highlighting the position of S. coccoides relative to the other type strains within the family Spirochaetaceae. The tree was inferred from 1,360 aligned characters [9,10] of the 16S rRNA gene sequence under the maximum likelihood criterion [11]. Rooting was done initially using the midpoint method [12] and then checked for its agreement with the current classification ( Table 1). The branches are scaled in terms of the expected number of substitutions per site. Numbers adjacent to the branches are support values from 500 ML bootstrap replicates [13] (left) and from 1,000 maximum parsimony bootstrap replicates [14] (right) if larger than 60% if. Lineages with type strain genome sequencing projects registered in GOLD [15] are labeled with one asterisk, those also listed as 'Complete and Published' with two asterisks (see [16][17][18][19], CP002696 for Treponema brennaborense, CP002903 for S. thermophila, and CP002868 for S. caldaria). Also, genomes that are finished but are missing a second asterisk are S. africana CP003282, S. pleomorpha CP003155 and S. globosa CP002541.  [20] and the NamesforLife database [21].

MIGS ID Property Term Evidence code
Current classification Domain Bacteria TAS [22] Phylum Spirochaetae TAS [23,24] Class Spirochaetes TAS [24,25] Order Spirochaetales TAS [26,27] Family Spirochaetaceae TAS [26,28] Genus Spirochaeta TAS [26,[29][30][31] Species a) Evidence codes -IDA: Inferred from Direct Assay; TAS: Traceable Author Statement (i.e., a direct report exists in the literature); NAS: Non-traceable Author Statement (i.e., not directly observed for the living, isolated sample, but based on a generally accepted property for the species, or anecdotal evidence). These evidence codes are from the Gene Ontology project [27]. If the evidence code is IDA, then the property should have been directly observed, for the purpose of this specific publication, for a live isolate by one of the authors, or an expert or reputable institution mentioned in the acknowledgements. Illumina reads were also used to correct potential base errors and increase consensus quality using a software Polisher developed at JGI [42]. The error rate of the completed genome sequence is less than 1 in 100,000. Together, the combination of the Illumina and 454 sequencing platforms provided 1,000.0 × coverage of the genome. The final assembly contained 137,682 pyrosequence and 58,694,953 Illumina reads.

Genome annotation
Genes were identified using Prodigal [43] as part of the Oak Ridge National Laboratory genome annotation pipeline, followed by a round of manual curation using the JGI GenePRIMP pipeline [44]. The predicted CDSs were translated and used to search the National Center for Biotechnology Information (NCBI) nonredundant database, UniProt, TIGR-Fam, Pfam, PRIAM, KEGG, COG, and InterPro databases. Additional gene prediction analysis and functional annotation was performed within the Integrated Microbial Genomes -Expert Review (IMG-ER) platform [45].

Genome properties
The genome consists of a 2,227,296 bp long chromosome with a G+C content of 50.6% (Table 3 and Figure 2). Of the 1,924 genes predicted, 1,866 were protein-coding genes, and 58 RNAs; 44 pseudogenes were also identified. The majority of the protein-coding genes (74.6%) were assigned with a putative function while the remaining ones were annotated as hypothetical proteins. The distribution of genes into COGs functional categories is presented in Table 4.

Insights from the genome sequence, and taxonomic conclusions for S. coccoides
Taxonomic interpretation for S. coccoides and neighboring species in the family Spirochaetaceae according to 16S rRNA data Based on its 16S rRNA sequence strain SPN1 T was placed into the genus Spirochaeta [1], although it lacks the typical spiral morphology and is nonmotile. SPN1 T showed highest similarity in 16S rRNA gene sequences to Spirochaeta sp. strain Buddy and Spirochaeta sp. strain Grapes [1], two spherical isolates that were not formally named at that time, but preliminarily named 'free-living pleomorphic spirochaetes' [4]. Recently, these isolates were placed into the novel genus Sphaerochaeta, and validly published as S. globosa and S. pleomorpha, respectively [4].
The phylogenetic tree shown in Figure 1 demonstrates that the current classification of the group suffers from a non-homogenous location of species featured as members of the genus Spirochaeta. Not only is Borrelia placed within Spirochaeta (without much branch support), but S. coccoides also appears as the sister group of Sphaerochaeta with maximum support. Support for a placement of S. caldaria, S. stenostrepta and S. zuelzerae more closely to Treponema than to the other Spirochaeta species (a topological arrangement that was observed earlier [46]) is also high and could only be considered a matter of rooting for the former two species (but note that the rooting is confirmed by a phylogenomic analysis described below and see the tree topology of the entire order Spirochaetales in [46,47]).  To measure phylogenetic conflict caused by the taxonomic classification in detail, we conducted both unconstrained heuristic searches for the best tree under the maximum likelihood (ML) [11] and maximum parsimony (MP) criterion [14] as well as searches constrained for the monophyly of all genera (for details of the data matrix see the caption of Figure 1). Our own re-implementation of CopyCat [48] in conjunction with AxPcoords and AxParafit [49] was used to determine those leaves (species) whose placement significantly deviated between the constrained and the unconstrained tree.
AxParafit was applied to the ML trees with 1,000 rounds of random permutations of the associations.
The ParaFit test was originally introduced for comparing host and parasite phylogenies [50], but can be applied to the comparison of all kinds of trees. In contrast to other measures for the comparison of trees, it includes a statistical test for whether individual leaves significantly contribute to the agreement between two trees (a p-value indicates how likely it is that this contribution is no more than random).
All other leaves apparently cause more conflict than agreement [50]. The rationale of comparing unconstrained trees with constrained trees inferred from the very same data is that the constraint might be in conflict with the original tree. In addition to assessing whether the trees are overall significantly different according to the data and a given optimality criterion in a paired-site test (see, e.g. chapter 21 in [51] for an in-depth description of such tests), the ParaFit test is a straightforward extension for assessing which leaves of the trees cause the conflict, if any. The best-known ML tree had a log likelihood of -16,001.40, whereas the best tree found under the constraint had a log likelihood of -16,322.98. The constrained tree was significantly worse than the globally best one in the Shimodaira-Hasegawa test as implemented in RAxML [11] (α = 0.01). The bestknown MP trees had a score of 3,105, whereas the best constrained trees found had a score of 3,260 and were significantly worse in the Kishino-Hasegawa test as implemented in PAUP* [14] (α < 0.0001). Accordingly, the current classification of the family as used by [3,46,47] is in significant conflict with the 16S rRNA data. Such discrepancies are not surprising in this group because many of the included taxa were described before 16S rRNA analysis could be applied [23,25,46], with Spirochaetaceae dating back to 1907 [28] and Spirochaeta even to 1835 [31]. Still uncultivated species and genera of Spirochaetales are described based on morphology alone, without depositing 16S rRNA sequences [52]. Table 5 shows the ParaFit test results obtained by comparing the unconstrained tree and the one obtained with the genus-based constraint. The largest conflict is caused by Spirochaeta aurantia, probably because of its placement close to Borrelia, followed by Sphaerochaeta and then by the other members of the main Spirochaeta group. To assess whether placing S. coccoides in Sphaerochaeta [4] and the other three Spirochaeta species that cause conflict in Treponema [46] would solve the problem, an according second constraint was created and used in phylogenetic analysis. The resulting ML tree had a log likelihood of -16,025.93 and was significantly worse than the best-known ML tree only for α = 0.05. The MP trees inferred under the second constraint had a score of 3,123 and were not significantly worse than the best-known MP trees. Table 5 also shows the ParaFit test results obtained by comparing the unconstrained tree and the one obtained with the second constraint. Apparently the conflict is largely resolved; the only remaining pvalue above 0.05 is the one for S. thermophilus, which is nevertheless only slightly above the chosen α-value (0.0539) and might become significant if more organisms were included [50]. Result (p-values) from the test of individual links with ParaFit for the species with an insignificant result (α = 0.05) in the first approach. The comparison was done between an unconstrained ML tree and the first, genus-based constraint (second column) or the second constraint, based on a revised classification of the group (third column). Note that with a single exception the phylogenetic conflict was resolved by assigning S. coccoides to Sphaerochaeta [4] and three other Spirochaeta species to Treponema [46].

Phylogenomic analyses
According to the results from 16S rRNA analysis and the whole-genome phylogenies described below, for a comparative analysis the genome sequences of S. globosa (GenBank CP002541) and S. pleomorpha (CP003155) [4], as well as the sequences of S. smaragdinae (GenBank CP002659) were used. An estimate of the overall similarity between S. coccoides, with both Sphaerochaeta species and S. smaragdinae was generated with the GGDC-Genome-to-Genome Distance Calculator [53,54]. This system calculates the distances by comparing the genomes to obtain HSPs (high-scoring segment pairs) and inferring distances from the set of formulas (1, HSP length / total length; 2, identities / HSP length; 3, identities / total length). Table 6 shows the results of the pairwise comparison. The comparison of S. coccoides with both Sphaerochaeta species revealed the highest scores using the GGDC. The comparison of S. coccoides with S. globosa and S. pleomorpha revealed that 4.5% and 3.9% of the average of genome length are covered with HSPs. The identity within the HSPs was 83.2% and 83.3%, respectively, whereas the identity over the whole genome was 3.7% and 3.3%, respectively. Lower similarity scores were observed in the comparison of S. coccoides with S. smaragdinae: only 1.2% of the average of either of the genome lengths are covered with HSPs. The identity within these HSPs was 84.6%, whereas the identity over the whole genome was only 1.0%. As expected, those distances relating HSP coverage (formula 1) and number of identical base pairs within HSPs to total genome length (formula 3) are higher between the S. coccoides and the Sphaerochaeta species than between S. coccoides and S. smaragdinae. That the distances relating the number of identical base pairs to total HSP length (formula 2) are different indicates that the genomic similarities between S. coccoides and S. smaragdinae are strongly restricted to more conserved sequences, a kind of saturation phenomenon [54].  59] with the OrthoMCL default parameters (an e-value threshold of 10 -5 and 2.0 as inflation parameter). OrthoMCL clusters containing inparalogs [57] were reduced as previously described [55] and aligned using MUSCLE version 3.7 under default settings [60]. The resulting alignments were filtered using RASCAL version 1.3.4 [61] and GBLOCKS version 0.91b [9] as in our earlier study [55]. Filtered alignments comprising at least four sequences were concatenated to form a supermatrix. As an extension of the approach in [55], the supermatrix was cleaned from relatively uninformative genes using MARE [62] under default values (except that deleting taxa was disallowed). Maximum-likelihood trees were inferred with RAxML [11] version 7.28 in conjunction with rapid bootstrapping and the bootstopping criterion [13] with subsequent search for the best tree. The best amino acid substitution model was determined beforehand by comparing the resulting log likelihoods on a maximum-parsimony starting tree. Maximumparsimony tree search was conducted with PAUP* version 4b10 [14] as previously described [55]. Standards in Genomic Sciences In addition to the supermatrix analysis, homologous sequences were determined using our own re-implementation of the TribeMCL algorithm [63] in conjunction with MCL [58,59], applying an e-value threshold of 10 -5 and an inflation parameter of 2.0. A gene-content (presence/absence) matrix was constructed, representing the occurrence of a gene of one genome within a cluster of homologs. Phylogenetic inference was done with the BINGAMMA model in RAxML and under maximum parsimony with PAUP*, other settings being as described above. The supermatrix comprised 2,408 genes and 696,696 characters before, 522 genes and 140,413 characters after cleaning with MARE. The selected model was PROTGAMMALGF; the resulting tree had a log likelihood of -2,172,190.75 and is shown in Figure 3. The best maximum-parsimony tree found had a length of 346,334 steps (not counting uninformative characters) and was topologically identical. The gene-content matrix comprised 11,131 characters and yielded a best tree with a log likelihood of -61,799.49 and a parsimony score of 10,229, respectively. Bootstrapping support values from all four methods applied are shown in Figure 3 if larger then 60%. The sister-group relationship of S. coccoides and Sphaerochaeta was unanimously supported by all methods, much like the placement of S. caldaria within Treponema. The trees differed however, regarding the support for the placement of Borrelia as sister group to all other ingroup taxa. For this reason, we assessed via long-branch extraction [64] whether this positioning could be caused by long-branch attraction [51] between Borrelia and the outgroup. Removal of Borrelia and subsequent phylogenetic inference yielded a maximum-parsimony tree with the same topology that would have been obtained by pruning Borrelia from the tree depicted in Figure 3. Removal of the outgroup from the alignment, however, yielded a maximum-parsimony tree in which Borrelia was placed as sister group of S. thermophila, supporting the long-branch attraction hypothesis (data not shown). The phylogenomic analysis thus confirms the 16S rRNA tree (Figure 1) regarding the paraphyly of Spirochaeta but, of course, based on much more characters. A first step to resolve this taxonomic problem is to assign S. coccoides to the genus Sphaerochaeta. Given that S. caldaria and some other species are situated within Treponema [46], and that Borrelia probably is placed within the remaining Spirochaeta species, further taxonomic changes will probably be necessary in the future. But apparently in addition to sampling more characters (by replacing 16S rRNA with genome sequences) sampling more taxa (by obtaining whole genomes from more type strains) might by necessary to obtain a natural classification of the spirochetes. Table 7 gives an overview of some morphological and physiological features of S. coccoides compared with the genus descriptions of Sphaerochaeta and Spirochaeta. The coccoid cell morphology, the cell size, the lack of motility as well as the products of fermentation support the reclassification of S. coccoides as a member of the genus Sphaerochaeta. S. coccoides is so close to the original description of the genus Sphaerochaeta that only its reported GC content needs to be modified. On the basis of the above mentioned physiological and phylogenetic characteristics of strain SPN1 T , its reclassification into the genus Sphaerochaeta is proposed. The inclusion of Sphaerochaeta in Spirochaetaceae also makes an emendation of the family necessary, as its previous description excludes features specifically found in Sphaerochaeta.

Emended description of the family Spirochaetaceae Swellengrebel 1907 (Spirochaetaceae Swellengrebel 1907 emend. Abt, Göker, Kyprides and Klenk)
The description of the family Spirochaetaceae is given by Swellengrebel 1907 [26,28]. Some species form coccoid cells, have no flagella and are not motile. Some do not have L-ornithine in the peptidoglycan.
Emended description of the genus Sphaerochaeta (Sphaerochaeta Ritalahti et al.

emend. Abt, Göker, Kyprides and Klenk)
The description of the genus Sphaerochaeta is as that given by Ritalahti et al. 2012 [4], with the following modification: DNA G+C content is 45-51 mol%.