Complete genome sequence and lifestyle of black-pigmented Corynebacterium aurimucosum ATCC 700975 (formerly C. nigricans CN-1) isolated from a vaginal swab of a woman with spontaneous abortion

Pyrosequencing and annotation of the C. aurimucosum ATCC 700975 genome

The complete genome sequence of the clinical isolate C. aurimucosum ATCC 700975 was determined by combining a whole-genome shotgun approach performed by pyrosequencing with a whole-genome scaffold generated by fosmid technology (Fig. 1A). A single run with the Genome Sequencer FLX System yielded 594,627 reads and a total number of 136,924,144 bases that were assembled into 73 large (≥ 500 bases) contigs. These assembled DNA sequences (2,736,233 bases) were combined with 215 Sanger reads performed on selected fosmid clones that covered the gaps between 70 chromosomal contigs. The resulting assembly of the DNA sequence data into a circular chromosome of 2,790,189 bp was consistent with the whole-genome scaffold deduced from 432 terminal insert sequences of randomly selected fosmid clones (Fig. 1B), corroborating the accuracy of the assembly process that was guided by the Consed program [15]. The gaps between the three remaining contigs (28,223 bases) were bridged by PCR, resulting in a circular contig of 29,037 bp in length. This assembled DNA sequence represents an endogenous plasmid of C. aurimucosum ATCC 700975, named pET44827 (Fig. 1C). The average gap size between the 70 chromosomal contigs was 1174 bp and 271 bp between the three plasmid contigs. Considering the final size of the C. aurimucosum ATCC 700975 genome, a 49-fold coverage was initially obtained by pyrosequencing. Relevant data of the C. aurimucosum ATCC 700975 genome project are summarized in Table 1.

Feature	Chromosome	Plasmid pET44827
Total size (bp)	2,790,189	29,037
G+C content (%)	60.6	53.3
No. of protein-coding sequences	2,531	20
No. of pseudogenes	4	8
Coding density (%)	88	64
Average gene length (bp)	973	928
No. of rRNAs	4 × 16S-23S-5S	0
No. of tRNAs	55	0
No. of transposase genes	85	5 (3 pseudogenes)
No. of CRISPR 1	98	0

¹Abbreviation: Clustered Regularly Interspaced Short Palindromic Repeats

The circular chromosome of C. aurimucosum ATCC 700975 has a size of 2,790,189 bp and a mean G+C content of 60.6% (Fig. 1A), which is slightly below the value of 63.7% that was previously determined by high-performance liquid chromatography [1]. Gene prediction and annotation of the genome sequence were performed with the GenDB system [16] and resulted in the detection of 2,531 protein-coding regions on the C. aurimucosum chromosome. On plasmid pET44827, twenty protein-coding sequences and eight pseudogenes were detected (Fig. 1C). Furthermore, 55 tRNA genes were found by the tRNAscan-SE program [17], and four rrn operons were identified on the leading strands of the C. aurimucosum chromosome, one on the right and three on the left replichore.

General architecture of the C. aurimucosum ATCC 700975 genome: moderate chromosomal reorganization in the main lineage of the genus Corynebacterium

The calculation of the G/C skew [(G-C)/(G+C)] [18] indicated a bi-directional replication mechanism of the C. aurimucosum ATCC 700975 chromosome (Fig. 1A). The origin of replication (oriC) was detected with Ori-Finder, a web-based tool for oriC prediction in bacterial genomes [19]. According to the presence and distribution of seven conserved DnaA boxes [TTATC(C/A)A(C/A)A], the oriC is located downstream of the dnaA (cauri_0001) coding region and has a computed length of 423 bp. The calculated G/C skew additionally indicated the presence of a putative dif region involved in replication termination [20] at the expected position of 180° from oriC, dividing the chromosome of C. aurimucosum ATCC 700975 in two replichores of similar size (Fig. 1A). For a more precise detection of a potential dif region, the distribution of the architecture imparting sequences G(A/T/C)GGGGGA and (T/C)GGGGGAG [21] was plotted on the leading and lagging strands of the C. aurimucosum ATCC 700975 chromosome (Fig. 2A). These octamers are present in 856 copies on the leading strands and in only 61 copies on the lagging strands of the C. aurimucosum ATCC 700975 chromosome. This characteristic distribution pattern of architecture imparting sequences revealed a putative dif region at around 1,425 kb of the chromosomal map (Fig. 2A). In accordance with this result, the respective DNA region of the C. aurimucosum ATCC 700975 chromosome contains a 28-bp sequence that showed striking similarity to the consensus sequence of actinobacterial dif sites [20] (Fig. 2B).

Synteny analysis by reciprocal best matches with BLASTP [22] revealed a conserved order of orthologous genes between the C. aurimucosum ATCC 700975 chromosome and other pathogenic corynebacteria, in particular with C. diphtheriae NCTC 13129 displaying two breakpoints of synteny at 1,298 kb and 1,537 kb of the C. aurimucosum chromosomal map (Fig. 3). The comparison between the gene order of C. aurimucosum ATCC 700975 and C. jeikeium K411, C. urealyticum DSM7109 and C. kroppenstedtii DSM44385 revealed additional breakpoints in the chromosomal organization (Fig. 3), which is consistent with the taxonomic position of the latter three species that are part of independent branches in the phylogenetic tree of the genus Corynebacterium [2]. The two breakpoints of synteny between the chromosomes of C. aurimucosum ATCC 700975 and C. diphtheriae NCTC 13129 are indicative of a symmetric inversion of a 239-kb chromosomal DNA segment (carrying the genes cauri_1172 to cauri_1381) around the terminus of replication. The detected inversion in the C. aurimucosum ATCC 700975 chromosome provides the first example of a moderate genetic reorganization in genomes of members from the main lineage of the genus Corynebacterium, since all hitherto completely sequenced genomes (C. diphtheriae NCTC 13129, C. glutamicum ATCC 13032, C. efficiens YS-314) displayed an overall synteny of their gene order [23].

Due to the symmetric inversion around the terminus of the C. aurimucosum ATCC 700975 chromosome, the preferential location of the respective genes on either the leading or lagging strands and their distance from the terminus region remained conserved after the chromosomal reorganization [24]. In general, 61% of the predicted C. aurimucosum ATCC 700975 genes are located on the leading strands of both replichores, while 39% are encoded on the lagging strands. It has been suggested that essentiality is the driving force for such type of gene-strand bias in bacterial genomes [25]. To detect candidate essential genes in the C. aurimucosum ATCC 700975 genome, the predicted gene repertoire was compared with that of C. glutamicum strain R [26], as genes shared by multiple genomes are more likely to be essential [27]. In addition, a high-density transposon mutagenesis approach in C. glutamicum R had already revealed 658 candidate essential genes in this strain, using complex medium and standard laboratory conditions for the selection of transposon mutants [28]. This number of candidate essential genes obviously represents an overestimate, at least due to polar effects caused by some transposon insertions [27]. The comparative genomics approach detected 352 coding regions of C. aurimucosum ATCC 700975 with high similarity to candidate essential genes from C. glutamicum R. Among this set of candidate essential genes, 78.8% are located on the leading strands of the C. aurimucosum ATCC 700975 chromosome, while 21.2% are encoded on the lagging strands, which is consistent with the aforementioned interpretation of gene-strand bias in bacterial genomes [25].

The annotation of the C. aurimucosum ATCC 700975 chromosome revealed the presence of two putative prophages. This observation is consistent with the lack of homology of the respective chromosomal region to the genomes of other pathogenic corynebacteria (Fig. 1A) and with a lack of synteny when compared with the genome sequence of C. diphtheriae NCTC 13129 (Fig. 3). The first prophage-like region (ϕCauriI) was detected adjacent to tRNA^Leu and tRNA^Gln genes of C. aurimucosum ATCC 700975 that may represent the integration site of ϕCauriI. This genome region has a size of about 20.7 kb and encodes 35 genes (cauri_1932 to cauri_1966). The second prophage-like region (ϕCauriII) is located directly adjacent to ϕCauriI and comprises 62 genes (cauri_1967 to cauri_2028) with a size of about 51.5 kb. Both prophage regions contain at their 3' ends genes encoding λ repressor-like transcription regulators (cauri_1965 and cauri_2027) and phage-related integrases (cauri_1966 and cauri_2028). Almost all genes of the prophage regions, with the exception of the regulatory genes, are encoded on the leading strand of the left replichore of the C. aurimucosum ATCC 700975 chromosome (Fig. 1A). The size difference between both prophage regions suggests that at least ϕCauriI is incomplete and a defective remnant of a formerly active corynephage.

Beside that, we detected with the CRISPRFinder program [29] an array of 98 clustered regularly interspaced short palindromic repeats (CRISPRs) and seven CRISPR-associated (cas) genes (cauri_0899 to cauri_0905) in the C. aurimucosum ATCC 700975 chromosome. The DNA repeats are 28 bp in length, separated by 33-bp spacers with variable nucleotide sequences and similar to repeat units of CRISPRs from C. jeikeium K411 [30], C. urealyticum DSM7109 [31] and Nocardia farcinica IFM 10152 [32]. CRISPR arrays and associated cas genes may provide acquired immunity against bacteriophages and other foreign genetic elements, with a specificity that is determined by sequence similarities between the spacers and foreign DNA sequences [33, 34].

Specific features of the carbohydrate metabolism of C. aurimucosum ATCC 700975: catabolism of aromatic amines and utilization of L-tartrate

The metabolic analysis of the C. aurimucosum ATCC 700975 genome was performed with the computer program CARMEN that provided an automatic mapping of genome annotations on manually curated metabolic pathway maps [35]. These data were combined with results from a comparative genomics approach performed with the bioinformatics tool EDGAR, resulting in the detection of orthologous genes in different corynebacterial genomes and their classification as core genes or species-specific genes, so-called singletons [36]. A comparison of the gene content of pathogenic corynebacteria revealed a putative core genome consisting of 1,048 genes that are common in C. aurimucosum, C. diphtheriae, C. jeikeium, C. urealyticum, and C. kroppenstedtii. This value is very similar to data deduced previously from the comparison of gene contents in pathogenic and non-pathogenic corynebacteria [26, 30]. The comparative genome analysis with EDGAR revealed also that C. aurimucosum ATCC 700975 contains 443 genes that are specific for this chromosome. The functional evaluation of the detected singletons provided characteristic features of the metabolism and lifestyle of C. aurimucosum ATCC 700975 that are described in more detail below. As the specific gene repertoire of a bacterium should in principle be shaped by the environmental conditions in the natural habitat and as we observed similarities to the gene repertoire of gut microbes, the annotation data point towards the human gut as a natural source for C. aurimucosum strains, which is consistent with the recent single report of a detection of this organism in a human fecal sample [37].

Analysis of the central carbohydrate metabolism of C. aurimucosum ATCC 700975 revealed the complete gene set for glycolysis and the pentose phosphate pathway (Fig. 4). The chromosome of C. aurimucosum ATCC 700975 provides a phosphoenolpyruvate carboxykinase (pck) and a fructose-1,6-bisphosphatase (pfkB) function for gluconeogenesis. The anaplerotic reaction of the tricarboxylic acid (TCA) cycle is mediated by pyruvate carboxylase (pyc). In contrast to other sequenced pathogenic corynebacteria that all lack the sucCD genes coding for succinyl-CoA synthetase, the TCA cycle of C. aurimucosum ATCC 700975 is complete (Fig. 4). It has been suggested that a putative succinyl-CoA:CoA transferase encoded by the cat1 gene can replace the missing succinyl-CoA synthetase activity in pathogenic corynebacteria [30, 38]. Besides sucCD (cauri_2063 and cauri_2062), the C. aurimucosum ATCC 700975 chromosome also carries the respective cat1 gene (cauri_2148). On the other hand, the C. aurimucosum ATCC 700975 genome lacks the aceAB genes for the glyoxylate bypass within the TCA cycle. The lack of aceAB was previously also observed in the genome sequence of C. diphtheriae NCTC 13129 [38].

The functional analysis of singletons in the C. aurimucosum ATCC 700975 genome sequence resulted in the detection of a distinct gene cluster for the uptake and degradation of aromatic amines (cauri_2491 to cauri_2501) (Fig. 4; Fig. 5A). C. aurimucosum ATCC 700975 can apparently take up aromatic amines by a specific permease encoded by hpaX. The subsequent transformation into ammonia and the corresponding aromatic acid is catalyzed by amine oxidase (hpaY) and phenylacetaldehyde dehydrogenase (padA). The resulting metabolite is converted into pyruvate and succinic semialdehyde via the homoprotocatechuate (HPC) meta-cleavage degradation route (Fig. 5A). The hpa gene cluster encoding the respective enzymes of this pathway consists of seven genes (hpaGLEDHI and gabD2). The gabD2 gene is required for the enzymatic dehydrogenation of succinic semialdehyde to succinate that can enter the TCA cycle (Fig. 4). The hpaF function necessary to convert 5-carboxymethyl-2-hydroxy-muconic acid (CHM) into 5-oxo-pent-3-ene-1,2,5-tricarboxylic acid (OPET) is encoded elsewhere in the C. aurimucosum ATCC 700975 chromosome (cauri_0890). Two regulatory genes (hpaL and cauri_2492) are apparently involved in controlling the expression of the HPC meta-cleavage pathway and the uptake and conversion of aromatic amines. The hpa gene cluster of C. aurimucosum ATCC 700975 is similar to a previously reported cluster for the uptake and degradation of aromatic acids in E. coli [39] (Fig. 5A). Aromatic compounds may constitute a carbon source for gut microbes, as they are not only abundant in soil and water, but are also present in the gastrointestinal tract, where they mostly derive from plant constituents. As the HPC 2,3-dioxygenase (HpaD) requires molecular oxygen, utilization of the HPC meta-cleavage pathway is restricted to distinct intestinal habitats, for instance ones in close proximity to epithelial cells, where oxygen molecules might pass from the blood to the microbes attached to the epithelium [40]. Accordingly, C. aurimucosum ATCC 700975 is probably able to utilize aromatic amines as carbon and energy sources when living in an aerobic environment, such as the peripheral areas of the human gut system.

Furthermore, the C. aurimucosum ATCC 700975 chromosome contains three singletons (cauri_2300 to cauri_2302) that are involved in the uptake and degradation of L-tartrate (Fig. 4). Uptake of L-tartrate is mediated by a tartrate/succinate antiporter encoded by yflS. The subsequent degradation to oxaloacetate is catalyzed by an (oxygen-labile) L-tartrate dehydratase (ttdAB). The respective gene cluster is homologous to previously described genes from E. coli [41, 42]. Tartaric acid is found in many fruits and vegetables and is also a product of microbial metabolism in the gut ecosystem that enhances the ability of the normal gut microflora to reduce the intestinal colonization by Salmonella species and other enteropathogens [43]. L-tartrate is the most widely distributed isomer and can be degraded under aerobic and anaerobic conditions by a variety of bacteria [41]. However, E. coli can utilize L-tartrate only as reducible substrate for supporting anaerobic growth on the oxidizable co-substrate glycerol [41]. The C. aurimucosum ATCC 700975 chromosome also encodes a transporter for the uptake of glycerol (glpF) as well as glycerol-3-phosphate (glpT), gluconate (gntP) and lactate (lldT) transporters (Fig. 4). Furthermore, five genes coding for components of the bacterial phosphotransferase system (PTS) and two additional enzyme II genes (manP, nagE) encoding proteins of the fructose, glucose and glucoside (sub)families of PTS permeases [44] were identified in the chromosome of C. aurimucosum ATCC 700975 (Fig. 4). This gene repertoire, associated with the central metabolism of C. aurimucosum ATCC 700975, suggested that this bacterium is very flexible in the utilization of different carbon and energy sources, under both aerobic and anaerobic growth conditions.

Specific features of the amino acid metabolism of C. aurimucosum ATCC 700975: L-histidine catabolism and utilization of selenocysteine in formate dehydrogenase

According to the genome annotation, all known pathways for the biosynthesis of standard proteinogenic amino acids are present in the C. aurimucosum ATCC 700975 chromosome. The detection of singletons with EDGAR revealed that C. aurimucosum ATCC 700975 encodes a complete histidine utilization (hut) pathway for the conversion of histidine to glutamate (Fig. 5B). Four gene products are directly involved in histidine catabolism: The first reaction is catalyzed by histidine ammonia-lyase (hutH; cauri_1023), followed by the conversion of the resulting urocanate to 4-imidazolone-5-propanoate via urocanate hydratase (hutU; cauri_1025). The next catabolic step generates N-formimidoyl-L-glutamate by imidazolonepropionase (hutI; cauri_1026) that is finally hydrolyzed into L-glutamate and formamide by formimidoylglutamase (hutG; cauri_1019). Similar gene regions are known from saprophytic bacteria, such as Pseudomonas species, that take up and utilize L-histidine when growing in the plant environment [45, 46] (Fig. 5B).

Regarding the proposal of C. aurimucosum ATCC 700975 as urogenital pathogen in pregnant women, it is interesting to note that the female histidine metabolism is affected during pregnancy, resulting in the excretion of large amounts of histidine in the urine [47, 48]. Histidine excretion is largely dependent on the composition of the diet and is generally higher during the second half of gravidity [47]. Although little information regarding the utilization by bacteria of L-histidine from human sources is available, the human gut microbe Fusobacterium varium was shown to use L-histidine as carbon source for growth [49]. According to the presence of a complete hut pathway, C. aurimucosum ATCC 700975 is apparently able to utilize L-histidine as carbon and/or nitrogen source, which might be beneficial for bacterial growth when colonizing the urogenital tract of pregnant women.

Moreover, the C. aurimucosum ATCC 700975 chromosome contains the selABD genes and the tRNA^Sec gene selC for the synthesis of the twenty-first proteinogenic amino acid selenocysteine (Sec) and its incorporation into selenoproteins (Fig. 5C). Selenocysteine is specifically incorporated into proteins by a cotranslational process that is directed by the UGA stop codon and termed recoding [50]. The first step during selenocysteine synthesis requires esterification of L-serine to the unique tRNA^Sec by L-seryl-tRNA synthetase (serS, cauri_2392). The seryl-tRNA^Sec is then converted to selenocysteyl-tRNA^Sec by selenocysteine synthetase (selA, cauri_0967), using selenophosphate as selenium donor, which in turn is synthesized by selenophosphate synthetase (selD, cauri_0966). The anticodon of the specific tRNA^Sec matches with an internal UGA stop codon in mRNAs from selenoprotein genes [51]. Furthermore, a specific elongation factor encoded by selB (cauri_0968) is required to recognize SECIS (selenocysteine insertion sequence) elements in the mRNA, such that the ribosome identifies the internal UGA stop codon as codon for selenocysteine. Translation of selenoprotein mRNAs then ends at the next in-frame stop codon to give a full-size selenoprotein [50, 51].

Interestingly, the fdnG and fdxH genes (cauri_0962 and cauri_0964) encoding subunits of a formate dehydrogenase are arranged in a cluster with coding regions for the synthesis and incorporation of selenocysteine (Fig. 5C). This gene order is highly conserved in bacterial genomes that code for the incorporation of selenocysteine into proteins. Formate dehydrogenase is a heterogeneous group of enzymes that catalyze the decomposition of formate to carbon dioxide and contribute to mixed acid fermentation. Some formate dehydrogenase enzymes can contain an intrinsic selenocysteine residue [52]. The fdnG coding region of C. aurimucosum ATCC 700975 includes an internal UGA stop codon, suggesting that a formate dehydrogenase of the selenoenzyme type is encoded by this DNA region. The C. aurimucosum ATCC 700975 chromosome carries an additional gene for a standard (thiol) formate dehydrogenase (fdhD, cauri_0426) that is located adjacent to a putative formate transporter gene (cauri_0425). The utilization of the selenoenzyme form of formate dehydrogenase may enhance the adaptability of C. aurimucosum ATCC 700975 to changing environments, due to the strikingly different pH-dependence when compared with thiol forms, with selenols being active at much lower pH [53]. This biochemical feature of selenoenzymes is remarkable when considering the lifestyle of C. aurimucosum ATCC 700975, since acidophilic lactobacilli create a low-pH environment in the human vagina by producing lactic acid [54].

Cell surface structures of C. aurimucosum ATCC 700975: synthesis of a SpaH-type pilus and potential in biofilm formation

The C. aurimucosum ATCC 700975 genome codes for 438 proteins containing predicted signal peptides for either the general secretory (sec) or the twin-arginine translocation (tat) pathway for protein secretion. Three secreted proteins of C. aurimucosum ATCC 700975 showed homology to subunits of the adhesive SpaH pilus from C. diphtheriae NCTC 13129 [55]. Adhesive pili are covalently anchored to the corynebacterial cell wall by a transpeptidylation mechanism, requiring a C-terminal sorting signal with a conserved LPXTG motif [56], and can mediate the adherence to host tissues and other bacterial cells [57]. The chromosome of C. aurimucosum ATCC 700975 contains two genes, srtD and srtE (cauri_0189 and cauri_0190), encoding putative sortases involved in the assembly of a SpaH-like pilus. The housekeeping sortase of C. aurimucosum ATCC 770975 is encoded elsewhere in the chromosome (srtB, cauri_2420). The srtDE genes of C. aurimucosum ATCC 700975 are part of a DNA region also comprising the spaHIG genes (cauri_0187, cauri_0188, cauri_0191) that encode surface-anchored proteins with C-terminal sorting motifs (Fig. 6A). These proteins showed similarity to fimbrial subunits from C. diphtheriae NCTC 13129 involved in the sortase-mediated formation of the SpaH pilus [55]. The SpaH protein represents the major pilin of this pilus type and, along with the minor pilin SpaI, forms the pilus shaft, whereas the SpaG protein is located at the tip of the pilus. According to the detected homology between the corynebacterial SpaHIG proteins, it is likely that C. aurimucosum ATCC 700975 expresses an adhesive pilus with structural similarity to the SpaH pilus of C. diphtheriae NCTC 13129 (Fig. 6B). This pilus structure can mediate the initial adhesion of C. aurimucosum ATCC 700975 to host tissues particularly during vaginal infection and/or colonization [58].

Further analysis of the genome sequence revealed that C. aurimucosum ATCC 700975 codes for a putative accumulation-associated protein (aap, cauri_1006). Bioinformatics analysis displayed five G5 domains within the amino acid sequence of Aap that can mediate the formation of biofilms by recognizing and binding to N-acetylglucosamines present in the peptidoglycan of the bacterial cell wall [59]. Moreover, the C-terminal segment of Aap contains a sorting motif (LANTG) for anchoring of the protein to the corynebacterial cell wall. The secretion mechanism of Aap remains unknown, as the protein is devoid of a typical signal peptide. The Aap protein of C. aurimucosum ATCC 700975 is also characterized by a C-terminal multiple repeat unit of six amino acids (PDEPGK), the number of which may vary among different isolates and during the course of an infection, as described for members of the biofilm-associated protein (Bap) family [60, 61]. Variations in the Aap protein sequence may help the bacterium to evade the immune response of the host, as observed in the case of the alpha C protein of group B streptococci [62]. The Aap protein of C. aurimucosum ATCC 700975 shares structural similarity with the biofilm-associated protein Bap of Staphylococcus aureus [63] and the accumulation-associated protein Aap of Staphylococcus epidermidis [64]. Both staphylococcal proteins are located on the cell surface and implicated in the formation of biofilms. Cell aggregation and biofilm formation by Aap requires proteolytic processing of the protein by endogenous or host-derived proteases [64]. In the latter case, host proteins can directly induce biofilm formation upon infection, thereby enabling the pathogen to evade the clearance by phagocytes. Both the Aap protein and the SpaH-type pilus apparently represent prominent protein factors on the surface of C. aurimucosum ATCC 700975 cells that are most likely involved in adhesion and colonization processes and in the formation of biofilms.

The gene repertoire of plasmid pET44827: synthesis of the characteristic black pigment of C. aurimucosum ATCC 700975

Assembly of the C. aurimucosum ATCC 700975 genome sequence revealed the presence of the endogenous plasmid pET44827 (Fig. 1C). The average G+C content of 53.3% is below the value of 60.6% determined for the C. aurimucosum ATCC 700975 chromosome, indicating the acquisition of pET44827 by horizontal gene transfer. Coding regions displaying amino acid sequence similarities with proteins of known functions and providing information about structural features of pET44827 and the physiological role of the deduced proteins are presented below.

Replication of pET44827 is presumably mediated by a replication initiator protein of the IncW type (repW) that showed 89% amino acid sequence identity to a homologous counterpart on the multiresistance plasmid pTP10 from C. striatum M82B [65]. This indicates that pET44827 replicates by the theta mechanism, in contrast to other corynebacterial plasmids belonging to the pNG2 family that replicate via the rolling-circle mode [66]. Additionally, pET44827 encodes a putative resolvase (gcrR) ensuring the stability of the plasmid during segregation. The putative resolvase of pET44827 contains a conserved N-terminal catalytic domain of the serine recombinase family and a C-terminal DNA binding domain [67]. However, the gcrR gene region lacks the typical DNA sequence repeats described for the homologous resolvase gene of pTP10 from C. striatum M82B [65]. Furthermore, pET44827 encodes a putative relaxase encoded by traA with 52% similarity to the relaxase of plasmid pREA400 from Rhodococcus erythropolis AN12 [68]. Two genes located upstream of traA were identified as parA and parB, showing similarity to class Ib plasmid partitioning genes present on the pNG2-family plasmid pTET3 from C. glutamicum LP-6 [69]. Therefore, pET44827 seems to be an unusual corynebacterial plasmid that contains characteristic features of the pNG2 family that are combined with a replication initiator protein of the IncW type.

The most prominent feature of pET44827 is a region composed of five genes, including orf18 coding for a putative non-ribosomal peptide synthetase (NRPS) (Fig. 1C). The G+C content of the NRPS gene region is about 47%, which is below the mean value detected for pET44827 (Table 1). As the NRPS gene region is flanked by a remnant of a transposase gene (tnpET23), it might have been integrated into the backbone of pET44827 by a transpositional recombination event. The deduced gene product of orf18 shares 52% similarity with the single module type NRPS BpsA from Streptomyces lavendulae [70] and 51% similarity with the NRPS IndC from the phytopathogen Erwinia chrysanthemi [71]. Both NRPSs catalyze the modular synthesis of the blue 3,3'-bipyridyl pigment indigoidine. The predicted NRPS of pET44827 comprises ten of the eleven highly conserved amino acid sequence motifs of NRPS proteins, only lacking a typical N-terminal condensation domain. Likewise, the IndC protein from E. chrysanthemi does not contain a condensation domain, suggesting that the final condensation step in pigment synthesis is catalyzed by other proteins [71]. NRPSs are key players in the synthesis of natural products and provide a modular assembly line by maintaining reaction intermediates covalently bound on the same phosphopantetheine prosthetic group [72]. NRPSs are synthesized as apoforms, followed by activation through phophospantetheinlytransfer to generate the holoenzyme. This activation might be mediated by the gene products of orf19 and orf20, encoding a protein with a phosphoribosyltransferase domain and a protein with homology to phosphoribosylanthranilate isomerase from Syntrophus aciditrophicus [73], respectively. Moreover, the NRPS gene region of plasmid pET44827 includes the coding region orf17 with deduced amino acid sequence similarity to IndA from Thermoanaerobacter tengcongensis, representing an uncharacterized protein involved in the pigment biosynthesis pathway of this species [74]. Upstream of the NRPS gene region, plasmid pET44827 encodes a putative transcriptional regulator of the MarR protein family (orf21), showing 53% similarity to the PecS regulatory protein from E. chrysanthemi [75]. Interestingly, the PecS protein is a global regulator of various virulence factors and controls the production of the extracellular blue pigment of E. chrysanthemi [75, 76]. According to this annotation of pET44827 genes, it is most likely that the characteristic black pigment of C. aurimucosum ATCC 700975 is synthesized by the concerted action of the NRPS encoded by orf18 and an unknown condensase function and that the black pigment molecule is relevant for the proposed pathogenicity of C. aurimucosum.

Bacterial strain and growth conditions

C. aurimucosum ATCC 700975 (DSM 44827, CCUG 48176, CIP 107436), formerly designated C. nigricans CN-1 [10], was obtained as a lyophilized culture from DSMZ (Braunschweig, Germany). The strain was originally isolated from the urogenital tract of a 34-year-old woman who experienced a spontaneous abortion during month six of pregnancy [10]. C. aurimucosum was routinely cultured on solid BYT complex medium [89] at 37°C. For the preparation of genomic DNA samples, C. aurimucosum was grown for 24 h in liquid BYT medium supplemented with 1.25% (w/v) glycine.

Sequencing of the C. aurimucosum ATCC 700975 genome

Genomic DNA of C. aurimucosum ATCC 700975 was purified from 20-ml aliquots of an overnight culture by an alkaline lysis procedure [90]. The published protocol was sligthly modified as follows: (i) 20-ml aliquots of C. aurimucosum cells were incubated in a 30 mg ml^-1 lysozyme solution at 37°C for 1 h. (ii) The harvested cells were lysed in 0.7 ml 10% (w/v) SDS solution at 37°C for 15 min. A total of 5 μg of purified genomic DNA was used for constructing a single-stranded template DNA (sstDNA) library. The concentration of the sstDNA library was quantified by using the Agilent RNA 6000 Nano Kit. Preparation of the sstDNA library and subsequent DNA sequencing were performed according to manufacturer's protocols (Roche Applied Science). Pyrosequencing of C. aurimucosum DNA was carried out with the Genome Sequencer FLX System (Roche Applied Science). The sequence reads were assembled with the GS Assembler Software (Version 1.1.02). Using the default cutoff of 500 bases for the classification of contigs, 73 large contigs (≥ 500 bases) and 40 small contigs were assembled to give a total size of 2,744,972 bases.

A fosmid library with a 7-fold coverage of the C. aurimucosum ATCC 700975 genome was constructed in E. coli EPI300 cells with the Copy Control Fosmid Library Production Kit and the pCC1FOS fosmid vector (Biozym Scientific). Terminal insert sequences from fosmid library clones were used to build a scaffold for the genome sequence assembly. The terminal insert sequences and the assembled genomic contigs obtained by pyrosequencing were combined by means of the Consed program [15]. To close the remaining gaps in the C. aurimucosum chromosome, additional Sanger reads were generated on purified DNA of selected "linking fosmids". The assembled contigs of plasmid pET44827 were bridged by PCR products obtained with Phusion DNA polymerase (Finnzymes) on genomic template DNA. All primers used in this study were synthesized by Eurofins MWG Operon.

Bioinformatics analysis and annotation of the genome sequence

To verify the assembly of the C. aurimucosum ATCC 700975 genome sequence, the paired-end sequence information obtained from fosmid clones was used to generate a genomic map with the BACCardI program [91]. Automated annotation of the C. aurimucosum genome sequence was carried out with the bacterial genome annotation platform GenDB [16] as described earlier [92], followed by manual curation. Proteins that are orthologous in completely sequenced pathogenic corynebacteria, C. diphtheriae, C. jeikeium, C. urealyticum, and C. kroppenstedtii [30, 31, 38, 92], were deduced by bidirectional best BLASTP hits [22]. Two proteins were considered as orthologs if BLASTP matches with at least 30% sequence identity and a minimum coverage of 50% of the query sequence length were detected in both directions. The orthologous pair of proteins have to represent additionally the best BLASTP hits for the respective query sequence in the target genome. Genes were added to the core genome of completely sequenced pathogenic corynebacteria if orthologs of the encoded proteins were detected in the complete set of selected genomes. If none of the BLAST hits met the criteria defined above in any other genome, the genes were classified as singletons (specific) for C. aurimucosum ATCC 700975. Gene content and comparative metabolic analyses were accomplished by the computer programs EDGAR and CARMEN [35, 36]. The annotated genome sequence of C. aurimucosum ATCC 700975 was submitted to the GenBank/EMBL database with accession numbers CP001601 (chromosome) and FM164414 (plasmid pET44827).