Porphyromonas gingivalis and Treponema denticola Exhibit Metabolic Symbioses

Porphyromonas gingivalis and Treponema denticola are strongly associated with chronic periodontitis. These bacteria have been co-localized in subgingival plaque and demonstrated to exhibit symbiosis in growth in vitro and synergistic virulence upon co-infection in animal models of disease. Here we show that during continuous co-culture a P. gingivalis:T. denticola cell ratio of 6∶1 was maintained with a respective increase of 54% and 30% in cell numbers when compared with mono-culture. Co-culture caused significant changes in global gene expression in both species with altered expression of 184 T. denticola and 134 P. gingivalis genes. P. gingivalis genes encoding a predicted thiamine biosynthesis pathway were up-regulated whilst genes involved in fatty acid biosynthesis were down-regulated. T. denticola genes encoding virulence factors including dentilisin and glycine catabolic pathways were significantly up-regulated during co-culture. Metabolic labeling using 13C-glycine showed that T. denticola rapidly metabolized this amino acid resulting in the production of acetate and lactate. P. gingivalis may be an important source of free glycine for T. denticola as mono-cultures of P. gingivalis and T. denticola were found to produce and consume free glycine, respectively; free glycine production by P. gingivalis was stimulated by T. denticola conditioned medium and glycine supplementation of T. denticola medium increased final cell density 1.7-fold. Collectively these data show P. gingivalis and T. denticola respond metabolically to the presence of each other with T. denticola displaying responses that help explain enhanced virulence of co-infections.


Introduction
Chronic periodontitis is an inflammatory disease of the supporting tissues of the teeth with a polymicrobial aetiology. However, whilst the concepts of the roles of particular oral bacterial species in disease have changed over the past two decades there is wide consensus that anaerobic, proteolytic, amino acid fermenting species including Porphyromonas gingivalis, Treponema denticola and Tannerella forsythia play a crucial role in initiation and/or progression of disease [1][2][3][4][5]. P. gingivalis has recently been proposed to be a ''keystone pathogen'' that through synergistic interactions aids the proliferation of other oral bacterial species resulting in the formation of a pathogenic polymicrobial plaque [3].
We have previously demonstrated in a longitudinal human study that the imminent progression of chronic periodontitis in patients on a maintenance program could be predicted by increases in the relative proportions of P. gingivalis and/or T. denticola in subgingival plaque above threshold levels [6]. This is consistent with other clinical studies demonstrating that P. gingivalis levels in subgingival plaque are predictive of human disease progression [7,8]. P. gingivalis and T. denticola are frequently found to co-exist in deep periodontal pockets and have been co-localized to the superfical layers of subgingival plaque as microcolonies adjacent to the pocket epithelium [1,[9][10][11][12][13], suggesting possible interbacterial interactions that might contribute towards disease [14,15]. When co-inoculated intra-orally in animal models of periodontitis P. gingivalis and T. denticola exhibit a synergistic pathogenesis [16][17][18]. These two bacteria display a symbiotic relationship in nutrient utilization and growth promotion in vitro [14,19] and bimodal coaggregation between P. gingivalis and T. denticola has been demonstrated [15,[20][21][22][23], which might explain their co-localization and aid synergistic biofilm production [24][25][26]. P. gingivalis and T. denticola responded to each other's presence in a polymicrobial biofilm by modulating the abundance of a range of proteins [26]. Together these data suggest there is an intimate relationship between these two species that has evolved to enhance their survival and virulence. However, the physiochemical interactions that result in the observed symbiotic and synergistic effects during P. gingivalis and T. denticola co-culture remain largely unknown.
In this study we used continuous co-culture to demonstrate that P. gingivalis and T. denticola symbiotically co-exist and that each bacterium adapts to the presence of the other by modulating gene expression, particularly those genes involved in metabolism and virulence. We show that the presence of P. gingivalis caused an upregulation of T. denticola glycine catabolism and that this amino acid supported the growth of T. denticola. T. denticola conditioned medium induced free glycine production by P. gingivalis implying intimate metabolic co-operativity between these species. The upregulation of T. denticola virulence factors in co-culture helps explain the synergistic virulence of P. gingivalis and T. denticola in animal models of disease.

Materials and Methods
Bacterial strains and culture conditions P. gingivalis strain W50 and T. denticola ATCC 35405 were obtained from the culture collection of the Oral Health Cooperative Research Centre, The University of Melbourne, and grown in oral bacterial growth medium (OBGM), that meets the growth requirements of both P. gingivalis and T. denticola [18,26,27]. Batch culture was in pre-reduced OBGM at 37uC in a MK3 anaerobic workstation (Don Whitley Scientific, Adelaide, Australia) with a gas composition of 5% CO 2 , 5% H 2 and 90% N 2 (BOC Gases, Wetherill Park, Australia). Bacterial cell density was monitored by measuring the absorbance at 650 nm (A 650 nm ).
To initiate the mono-species continuous culture, 300 mL of batch-grown inoculum in exponential growth phase (A 650 nm of ,0.6 for P. gingivalis or A 650 nm of ,0.2 for T. denticola) was mixed with 600 mL of fresh OBGM and transferred to a Bioflo 110 Modulator Benchtop Fermentor (New Brunswick Scientific, NJ, USA). The culture was maintained at 37uC, continuously agitated at 50 rpm and gassed with a constant stream of anaerobic gas (10% CO 2 in N 2 ) (BOC Gases). After 24 h, medium flow was commenced with OBGM pumped into the fermentor maintaining a working volume of 900 mL, with a flow rate of 39 mL h 21 , giving a mean generation time of 15.75 h. Samples were collected after the A 650 nm of the culture remained stable for a period of ten generations (158 h).
To establish co-culture P. gingivalis was inoculated into established T. denticola mono-species continuous cultures. After the T. denticola samples were harvested from a mono-culture chemostat, co-culture was established by replacing 300 mL of T. denticola culture with batch-grown exponential phase P. gingivalis (A 650 nm ,0.6). The medium flow to the fermentor was stopped for 12 h to enable establishment of co-culture, after which medium flow was resumed at 39 mL h 21 . Three independent biological replicates each of P. gingivalis and T. denticola mono-and co-cultures were grown.

Bacterial enumeration
Mono-culture bacterial cell density was determined by correlating A 650 nm to a standard curve [28] whilst P. gingivalis and T. denticola cells numbers in co-culture were determined by quantitative real-time PCR (qPCR) as described previously [6,26,28].

DNA microarray slides
The 60-mer oligonucleotide probes representing predicted open reading frames (ORFs) of the P. gingivalis and T. denticola genomes to be used for microarray slide preparation were designed using OligoArray 2.l [29] and using the bioinformatic services of Illumina Inc. (CA, USA). The difference in GC content of P. gingivalis (48%) and T. denticola (38%) genomes was exploited in probe design. BLAST analysis demonstrated that probes from one organism had less than 80% similarity to that from the other organism to minimize the chances of cross-species hybridization. The custom made T. denticola ATCC 35405 microarray contained 2518 probes representing 89% of the predicted ORFs in the T. denticola genome as described previously in Mitchell et al. [30]. The P. gingivalis probe set design was based upon the P. gingivalis W83 genome sequence [31] and ORF predictions available through The J. Craig Venter Institute (www.jcvi.org), with additional P. gingivalis ORFs predicted by the Los Alamos National Laboratory Oralgen project (www.oralgen.lanl.gov). The final set of 1977 probes corresponded to 96% of the predicted P. gingivalis ORFs. The full complement of probes for both genomes were printed twice each onto Corning UltraGAPs coated slides using a Virtek Microarray spotter by the Australian Genome Research Facility (Melbourne, Australia). Microarray Sample Pool (MSP) control probes were also included to aid intensity-dependent normalisation [32].
Total RNA extraction and purification RNA was isolated using the GenElute Total RNA Purification Kit (Sigma Aldrich, Castle Hill, Australia) and treated with DNase using the Turbo DNA-free kit (Ambion, TX, USA). The amount, integrity and purity of the RNAs were assessed using the Experion Automated Electrophoresis System (Bio-Rad, CA, USA). The absence of genomic DNA contamination from P. gingivalis and T. denticola in the RNA sample was determined by performing PCR on RNA samples using TDE0762 (forward: GGCTCCGAATC-AAAACGATA, reverse: CTATCGACTCCCCGTTTTCA) and PG0719 (forward: GCATTGCAGCATAGCGAATA, reverse: G-CCGATGGAAAAAGTGTGTT) primer pairs respectively while the respective genomic DNAs were used as positive controls.

Microarray analysis
Aliquots of P. gingivalis and T. denticola (100 mL) were harvested from 3 mono-cultures and from 3 co-cultures, with technical replicates harvested 2 days apart. qPCR showed that there were six times as many P. gingivalis as T. denticola in co-culture (vide infra) therefore 6:1 equivalents of P. gingivalis:T. denticola mono-culture RNAs were mixed prior to cDNA synthesis (vide infra) for use in comparison with the co-culture cDNA. Labeled cDNA hybridization targets were produced using either 5 mg of genomic DNA or 6 mg of purified total RNA template, 5 mg of random hexamers (Life Technologies, MD, USA) and aminoallyl dUTP nucleotides incorporated during cDNA synthesis. The genomic cDNA was

Author Summary
Unlike the traditional view that most diseases are caused by infection with a single bacterial species, some chronic diseases including periodontitis result from the perturbation of the natural microbiota and the proliferation of a number of opportunistic pathogens. Both Porphyromonas gingivalis and Treponema denticola have been associated with the progression and severity of chronic periodontitis and have been shown to display synergistic virulence in animal models. However, the underlying mechanisms to these observations are unclear. Here we demonstrate that these two bacteria grow synergistically in continuous coculture and modify their gene expression. The expression of T. denticola genes encoding known virulence factors and enzymes involved in the uptake and metabolism of the amino acid glycine was up-regulated in co-culture. T. denticola stimulated the proteolytic P. gingivalis to produce free glycine, which T. denticola used as a major carbon source. Our study shows P. gingivalis and T. denticola co-operate metabolically and this helps to explain their synergistic virulence in animal models and their intimate association in vivo.
produced using Platinum Taq DNA polymerase (Life Technologies) whilst RNA was reverse transcribed using Superscript III Reverse Transcriptase (Life Technologies). cDNA were purified using QlAquick columns (Qiagen, CA, USA) and labeled with monoreactive Cy3 or Cy5 dye (40 nmol) (GE Healthcare Lifesciences, Quebec, Canada). Equal numbers of samples from mono-and co-culture cDNAs were labeled with Cy3 and Cy5 respectively and in reverse combination in order to accommodate for different Cy3 and Cy5 labeling efficiencies. To ensure the specificity of the probes, the dual-genome array was probed with P. gingivalis or T. denticola labeled cDNA. Five T. denticola probes (TDE0780, TDE1033, TDE1804, TDE2113 and TDE2213) were found to cross-hybridize to P. gingivalis DNA and six P. gingivalis probes (PG1340, PG1473, PG1525, PG1666, PG1731 and PG2175) were found to cross-hybridize to T. denticola DNA. The expression of these genes was excluded from the subsequent transcriptomic analysis. The microarray hybridization and scanning were conducted essentially as described previously [30,33] but with 49% formamide in the hybridization buffer and use of 46uC as the hybridization temperature.

Data analysis
Microarray background subtraction and data analysis were conducted in R statistical environment using GenePix Pro 6.0 and LIMMA as described previously [30,34]. The Benjamini-Hochberg method was used to control the false discovery rate to correct for multiple testing [35]. Genes with a fold change equal to or above 1.4 and an adjusted p#0.05 were considered to be significantly differentially regulated. Annotation and putative protein functions were based on the National Center for Biotechnology Information Refseq database and Clusters of Orthologous Groups of protein (COG) database [36]. Operon predictions, COG assignments and COG functional categories were obtained from the Microbes Online database [37]. Microarray data are available in the ArrayExpress database (www.ebi.ac. uk/arrayexpress) under accession number E-MTAB-2214 (P. gingivalis) and E-MTAB-2257 (T. denticola).

Validation of microarray results using quantitative reverse transcription PCR (qRT-PCR)
Differential expression of selected genes was validated by qRT-PCR of cDNAs using SYBR Green-based detection on a Rotor Gene 3000 system (Corbett Research, Sydney, Australia) as reported previously [30]. To normalize the amount of mRNA in each reaction, a pool of six reference genes TDE2535, TDE0872, TDE1208, TDE1999, TDE1226 and TDE0002 [30] was assessed for expression stability using the geNorm program [38]. Analysis indicated that the geometric mean of the three most stable genes TDE2535, TDE1208 and TDE0872 with average pair-wise variations of 0.146 was sufficient to calculate a normalization factor for each sample. Primers for the validation gene set  TDE1624, TDE1625, TDE1626, TDE1627, TDE1259, TDE2119,  TDE2120, TDE0405, TDE0762, TDE1669, TDE0387, TDE0627 and TDE0832 were designed using Primer3 [39] (Table S1).

Scanning electron microscopy
Co-cultures from continuous culture were collected then processed and imaged using a Philips XL30 field-emission scanning electron microscope at a voltage of 2 kV as described previously [30].

Analysis of glycine concentration
Batch cultures of P. gingivalis (50 mL) with a starting cell density of 1610 7 cells were grown using a) OBGM, b) OBGM supplemented 1:1 with phosphate buffered saline (PBS) (OBGM/PBS) and c) OBGM supplemented 1:1 with cell-free T. denticola conditioned growth medium (OBGM/T. denticola conditioned medium). T. denticola conditioned medium was prepared by filtering (0.1 mm) a 7-day culture of T. denticola grown in OBGM. Aliquots of the culture medium (30 mL) were collected and immediately diluted with 70 mL of deionized water, 300 mL of CHCl 3 and 100 mL of CH 3 OH. The extracted medium was snapfrozen on liquid nitrogen and stored at 280uC for later gas chromatography-mass spectrometry (GC-MS) analysis [40] (Protocol S1). A glycine standard curve was obtained by the addition of various amounts of glycine to T. denticola conditioned medium (which had no detectable free glycine) in preference to use of pure glycine standards to account for matrix effects from the medium. The difference in free glycine levels relative to that at t = 0 h for each replicate was expressed as a function of P. gingivalis cell numbers in different media. A regression line was fitted using a linear mixed modelling approach, which allowed the regression lines to be fitted with random slopes by considering the fixed effects for the different growth medium and time interaction (SPSS version 20, IBM, IL, USA). For the determination of the total glycine content of cell-free media samples were freeze-dried and hydrolyzed at 150uC for 2 h in the presence of 6 M HCl containing phenol (1% v/v).

Isotope-labeled glycine studies
[U-13 C]glycine (Cambridge Isotope Laboratories, MA, USA) was added to a final concentration of 5 mM to a 24 h T. denticola culture in OBGM. Bacterial culture (1.5 mL) was collected every 24 h for 8 days and passed through a 0.22 mM filter to obtain a cell-free fraction. Samples were kept at 280uC until further analysis. Sample preparation for 13 C-nuclear magnetic resonance (NMR) spectroscopy and spectra analyses were conducted by Metabolomics Australia (Melbourne, Australia) as previously published [40].
Thiamine pyrophosphate (TPP) production by P. gingivalis Escherichia coli JRG902 [41] and E. coli JW3957-1, a TPP auxotrophic strain (DthiE764::kan), were obtained from the Keio collection [42]. E. coli strains were grown in M9 medium [43] with different combinations of TPP (5.88 nM) and/or cell-free P. gingivalis conditioned medium that was obtained by growing P. gingivalis in OMIZ-M/TD [44] without TPP and filtering (0.2 mM) the culture. Growth was measured by A 650 nm .

P. gingivalis and T. denticola continuous culture
Mono-and co-cultures of P. gingivalis and T. denticola were established in continuous culture using OBGM. Three independent continuous mono-cultures of P. gingivalis entered steady state 3 days after establishment with an average absorbance (A 650 nm ) of 1.8660.14 equating to 3.960.3610 9 cells mL 21 . The three T. denticola mono-cultures also reached steady state approximately 8 days after inoculation with an average A 650 nm of 0.2360.02 equating to 7.760.7610 8 cells mL 21 (Figure 1). Co-cultures of P. gingivalis and T. denticola entered steady-state after 4 days with an average A 650 nm of 1.7360.09 (Figure 1). There were 6.060.7610 9 P. gingivalis cells mL 21 and 1.060.1610 9 T. denticola cells mL 21 under steady state conditions for the three biological replicates as determined by qPCR. Thus cell densities of both P. gingivalis and T. denticola increased significantly, by 54% and 30% (p,0.01) respectively, in co-culture compared with mono-culture.
Interestingly, the 50% increase in total bacterial cell number during co-culture was not reflected by a concomitant increase in absorbance. SEM analysis of a co-culture showed that there was considerable coaggregation between the species with P. gingivalis adhering along the entire length of T. denticola cells, which would explain the lower culture A 650 nm than would be expected for this density of discrete cells (Figure 2).

Differential gene expression in co-culture
The expression of 184 T. denticola genes (6.6% of the genome) and 134 P. gingivalis genes (7.4% of the genome) was differentially regulated (adjusted p#0.05 and fold change of $1.4) in co-culture in comparison with mono-culture. Many of these genes were predicted to be polycistronic, with 33 predicted T. denticola and 16 predicted P. gingivalis operons containing two or more differentially co-regulated genes. The direction of change of expression was consistent within the predicted operons, with no single operon containing both up-regulated and down-regulated genes (Table  S2; Table S3). Differentially expressed genes were sorted into functional categories on the basis of clusters of orthologous groups (COG) [45]. The majority of T. denticola (103/184) and P. gingivalis (80/134) differentially expressed genes encoded proteins that belong to category R (general function prediction), category S (function unknown) or the not assigned category (Table 1).  . Scanning electron micrograph of P. gingivalis and T. denticola grown in continuous culture. Co-culture was collected, fixed on a coverslip, dehydrated, covered with colloidal silver, gold-coated and imaged using a Philips XL30 field-emission scanning electron microscope. Electron micrographs showed that P. gingivalis and T. denticola coaggregated. T. denticola is a long helical shaped spirochete with an average length of 5 to 20 mm. P. gingivalis is a coccobacillus with an average diameter of 1 mm. Putative T. denticola outer sheath vesicles and/or P. gingivalis outer membrane vesicles are indicated by arrows along the length of T. denticola. doi:10.1371/journal.ppat.1003955.g002

T. denticola differentially expressed genes
Thirty-four differentially expressed T. denticola genes were assigned to COGs related to metabolism, with 15 of these clustering in category E (amino acid transport and metabolism), representing 15% of the total genes assigned to this COG (Table 1). TDE0392 and TDE0389 which encode the putative Fe-S-dependent b subunits (HgdCA, HgdB) of (R)-2-hydroxyglutaryl-CoA dehydratase, an iron-sulfur cluster (4Fe-4S)-dependent enzyme involved in the fermentation of glutamate, as well as TDE0387 which encodes its 4Fe-4S-dependent activator HgdC, were down-regulated during co-culture. In addition, the gene encoding carbamate kinase (TDE2476) was down-regulated suggesting reduced glutamate catabolism by T. denticola when co-cultured with P. gingivalis. In contrast genes encoding enzymes involved in the glycine cleavage system (GcvP1, GcvP2 and GcvH), the glycine reductase system (Protein B2) and oligopeptide/dipeptide/amino acid transporters (OppA, TDE1067, TDE0985) were up-regulated suggesting an increased glycine catabolism by T. denticola.
The increased expression of genes encoding components of the glycine cleavage and glycine reductase systems was confirmed by qRT-PCR using TDE2535, TDE0872 and TDE1208 as stable reference genes. There was a significant correlation between the expression ratios determined by both microarray and qRT-PCR (R 2 = 0.9839) ( Figure S1). This comparison revealed a slight compression of the gene expression data from the DNA microarray analysis and gave a higher up-regulation of components of the glycine cleavage and glycine reductase systems with a .1.4 fold up-regulation of TDE1627, the T-protein of the glycine cleavage system and TDE2120, the GrdE2 of the glycine reductase system ( Table 2).

P. gingivalis differentially expressed genes
Co-culture with T. denticola had no effect on the transcription of P. gingivalis genes belonging to categories Q (secondary metabolites biosynthesis, transport and catabolism) and D (cell cycle control, cell division, chromosome partitioning) (Table 1). Furthermore, little effect on the transcription of genes in categories M (cell wall/ membrane/envelope biogenesis), C (energy production and conversion), G (carbohydrate transport and metabolism), E (amino acid transport and metabolism) and F (nucleotide transport and metabolism) was observed. Six differentially expressed genes in category H (coenzyme transport and metabolism) were upregulated, four of which (thiH, thiG, thiE/D, thiS) occur in a predicted five gene operon (PG2107-11) suggesting increased thiamine biosynthesis during co-culture. In addition PG2010 (thiC) was also significantly up-regulated (adjusted p = 0.01) but did not meet the 1.4-fold cut-off criterion. The putative thiamine transporter, PnuT (PG1898) was significantly up-regulated 1.5 fold (Table S3). Three genes involved in the initial stages of fatty acid biosynthesis fabG, fabF, acpP were down-regulated (Table S3).

Free glycine use by T. denticola
The increased transcription of genes encoding enzymes in glycine metabolism in T. denticola when co-cultured with P. gingivalis prompted us to investigate T. denticola glycine metabolism further. Suspension of T. denticola in OBGM containing a starting concentration of 1.46 mM glycine, resulted in the rapid depletion of this amino acid over 72 h, coinciding with entry into stationary growth (Figure 3a). Further supplementation of the OBGM to 10 mM glycine enhanced T. denticola growth, resulting in a 1.75fold increase in final cell density. Addition of glycine to a stationary T. denticola culture caused the resumption of bacterial growth (Figure 3b). These data show that T. denticola consumed free glycine and that glycine availability had a significant impact on T. denticola growth.

Metabolic fate of free glycine
Direct evidence for glycine catabolism in T. denticola was provided by metabolic labeling with 5 mM [U-13 C]glycine. [U-13 C]glycine was added to a 24 h batch-grown T. denticola culture and consumption of 13 C-glycine and production of 13 Clabeled end-products determined by 13 C-NMR analysis of the culture medium ( Figure S2). [U-13 C]glycine was completely consumed by 144 h, with production of 13 C-acetate (3.67 mM) and 13 C-lactate (1 mM) (Figure 4). The major isotopomers of acetate and lactate were uniformly labeled, although low levels of [1-13 C]acetate and/or [2-13 C]acetate, were also detected (data not shown).
Catabolism of 13 C-glycine was also associated with the production of H 13 CO 3 . The yields of H 13 CO 3 (0.17 mM) are likely to be an underestimate due to equilibration with the CO 2 enriched atmosphere above the medium. A small amount of 13 Clabeled alanine was also produced with ,0.14 mM being detected at the 168 h time point.
T. denticola conditioned medium stimulates free glycine production by P. gingivalis Increased expression of T. denticola genes encoding glycine catabolic pathways during co-culture suggested that there may be increased glycine availability. We therefore determined if P. gingivalis Table 2. Expression of T. denticola genes encoding enzymes involved in glycine or glycine-related metabolism during co-culture with P. gingivalis. growth could provide this additional glycine and if T. denticola could stimulate P. gingivalis glycine production. When P. gingivalis was grown in OBGM there was a small increase in free glycine ( Figure 5) while no increase in free glycine was evident in uninoculated OBGM over the same time period (data not shown). When P. gingivalis was grown in OBGM/T. denticola conditioned medium free glycine increased from 0.7560.04 mM at time 0 h to 2.2660.66 mM after 46 h, a significant difference of 1.51 mM (p,0.01) whereas the free glycine in the OBGM/PBS control culture increased only 0.2660.04 mM (p,0.01) ( Figure 5) after 46 h. In contrast free glycine content was unchanged in the uninoculated OBGM/T. denticola conditioned medium control ( Figure 5). To account for the differing number of P. gingivalis cells in the different media over time and the influence of this on free glycine generation, the change in glycine concentration was expressed as a function of P. gingivalis cell number. A regression line was fitted using a linear mixed modelling approach with the slope representing glycine production per 10 9 P. gingivalis cells (Figure 6). Glycine production by P. gingivalis grown in OBGM/PBS was 0.17160.012 mmole glycine/10 9 cells which was not statistically different from that determined in OBGM at 0.16460.020 mmole glycine/10 9 cells (p.0.10) ( Figure 6). However, free glycine production by P. gingivalis in OBGM/T. denticola conditioned medium was 0.54960.090 mmole/10 9 cells, which was more than three times that observed in OBGM or OBGM/PBS (p,0.01).    To determine the source of this increased free glycine the free and total glycine concentrations were determined prior to and 48 h after P. gingivalis inoculation in OBGM/T. denticola conditioned and OBGM/PBS media. Peptide-bound glycine (totalfree) decreased by 1.26460.143 mmole/10 9 cells in the OBGM/T. denticola conditioned medium, significantly (p,0.01) more than the 0.73360.056 mmole/10 9 cells decrease in OBGM/PBS. These data indicate a higher rate of peptide hydrolysis and release of free glycine by P. gingivalis in the OBGM/T. denticola conditioned medium. Total glycine decreased by 0.76560.090 mmole/10 9 cells in the OBGM/T. denticola conditioned medium, compared with the 0.58360.030 mmole/10 9 cells decrease in OBGM/PBS, indicating a slightly higher rate of uptake by P. gingivalis in the OBGM/T. denticola conditioned medium.

P. gingivalis TPP production
The increased transcription of P. gingivalis genes during coculture that encoded thiamine biosynthesis and transport-related proteins prompted us to examine whether P. gingivalis produces excess thiamine that could be used by T. denticola, a thiamine auxotroph. Thiamine pyrophosphate (TPP) is a micronutrient that is required in extremely low concentrations for bacterial growth that are difficult to detect biochemically. We therefore used an E. coli auxotrophic strain, JW3957-1 to determine excess thiamine production and release by P. gingivalis. E. coli JW3957-1 was unable to grow in M9 minimal medium supplemented with uninoculated P. gingivalis growth medium unless it was supplemented with 5.88 nM TPP. In contrast the E. coli parent strain JRG902 was able to grow to a similar cell density with or without 5.88 nM TPP (Figure 7a). The addition of P. gingivalis cell-free spent medium caused inhibition of E. coli JRG902 growth by an undefined mechanism that was independent of TPP addition (Figure 7b). However the addition of P. gingivalis cell-free spent medium did enable limited growth of E. coli JW3957-1, indicating that there was some available TPP in the medium that had been produced and released by P. gingivalis (Figure 7b). Addition of 5.88 nM TPP resulted in a similar final cell density of both E. coli JW3957-1 and JRG902.

Discussion
P. gingivalis and T. denticola are frequently found together in subgingival plaque samples taken from diseased periodontal sites [1,6,10] as spatially co-localized micro-colonies on the surface of the plaque adjacent to the host pocket epithelium [12,13]. This implies a strong ecological relationship and potential interactions that may contribute to the progression of chronic periodontitis. In Figure 6. Free glycine production during P. gingivalis growth. The difference in the amount of free glycine relative to that at t = 0 h as a function of P. gingivalis cell numbers in a) OBGM/PBS, b) OBGM and c) OBGM/T. denticola conditioned medium. A regression line was fitted using a linear mixed modelling approach. The slope represents the amount of glycine produced/10 9 P. gingivalis cells. doi:10.1371/journal.ppat.1003955.g006 Figure 7. Excess thiamine pyrophosphate production by P. gingivalis. The E. coli TPP auxotrophic strain JW3957-1 (black shading) and the parent strain JRG902 (white shading) were cultured in M9 growth medium (that lacks thiamine) supplemented with either (a) uninoculated P. gingivalis medium (that lacks thiamine) or (b) cell-free P. gingivalis spent medium. The bacterium was cultured with or without TPP addition (5.88 nM). doi:10.1371/journal.ppat.1003955.g007 this study, we showed that the sum of the interactions (including competition and cooperation) between P. gingivalis and T. denticola resulted in sustained interspecies growth symbiosis in continuous culture over an extended time frame. In addition, the two species reached a reproducible steady state with a cell density significantly higher than in mono-culture that strongly suggests metabolic synergy.
A dual species transcriptome analysis determined that T. denticola and P. gingivalis responded to each other in co-culture by altering the expression of a substantial proportion of genes, notably those involved in T. denticola metabolism and virulence. The shift in metabolism was evidenced by altered expression in genes encoding proteins involved in iron acquisition, utilization and storage by each species, altered glutamate and glycine catabolism by T. denticola and changes in P. gingivalis fatty acid and thiamine pyrophosphate synthesis. This suggests that there may be metabolites produced by each species that enables the other to bypass or alter specific metabolic processes. The upregulation of the expression of P. gingivalis genes encoding thiamine pyrophosphate biosynthesis indicated an increased production of TPP by P. gingivalis. This may be a result of competition with T. denticola, which is auxotrophic for TPP, or may indicate that that there is some cross-feeding of TPP from P. gingivalis to T. denticola. Due to the difficulties of biochemically measuring nanomolar concentrations of TPP we used an E. coli TPP auxotrophic strain to demonstrate that P. gingivalis produces free TPP that can be utilized by other bacterial species. This indicates that there is a likelihood that T. denticola benefits from P. gingivalis TPP biosynthesis and release in co-culture.
Genes encoding proteins involved in iron storage, heme acquisition and thioredoxin were down-regulated in P. gingivalis co-cultured with T. denticola indicating that P. gingivalis experienced a metabolic shift as a result of co-culture. P. gingivalis is auxotrophic for porphyrin, therefore also for heme and cobalamins as it lacks key enzymes in the early steps of porphyrin biosynthesis [31]. The P. gingivalis gene hmuY (PG1551) that encodes an outer-membrane hemin binding protein important in heme acquisition was the most down-regulated P. gingivalis gene during co-culture with T. denticola, supporting our recent finding of a significant decrease in HmuY abundance in a polymicrobial biofilm containing P. gingivalis and T. denticola [26]. Succinate produced by T. denticola has been reported to alleviate the P. gingivalis requirement for heme during growth in heme-limited medium [14,46].
In addition fatty acid cross-feeding has previously been demonstrated between P. gingivalis and T. denticola [15] which is consistent with the down-regulation of genes encoding enzymes participating in the initial stage of fatty acid synthesis found in our study. Hence, these results suggest that the presence of T. denticola helps P. gingivalis reduce energy consuming processes which may explain the increase in cell biomass of P. gingivalis grown in the presence of T. denticola.
The coaggregation of T. denticola with P. gingivalis in co-culture as shown by SEM would assist T. denticola in uptake of nutrients produced by P. gingivalis, and vice versa. In the polymicrobial biota of subgingival dental plaque, which is subject to the flow of gingival crevicular fluid, the ability of T. denticola to adhere to other bacteria and to stimulate production of metabolites such as glycine by other species to provide energy and to compensate for it auxotrophies [44,47] would be of significant benefit to survival and thus virulence. The motility of T. denticola and chemotactic responses would also be significant to its survival. Decreased expression of genes encoding MCP chemotaxis receptors by T. denticola in co-culture with P. gingivalis relative to mono-culture indicates that several substrates, possibly glycine and TPP are in greater abundance in the co-culture such that chemotaxis to a more prefered environment is of lesser importance.
Genes involved in T. denticola glutamate metabolism were downregulated in co-culture. This potential shift in catabolism may be related to reduced FeS-cofactor biosynthesis and competition as P. gingivalis has a preference for glutamate and aspartate [48,49]. Significant up-regulation in the expression of T. denticola genes encoding peptidases and enzymes involved in glycine catabolism in co-culture relative to mono-culture was also observed. T. denticola is predicted to have an alanine/glycine cation symporter (TDE1259), a complete glycine cleavage system and the glycine reductase system [47]. Glycine is utilized as an important energy and carbon source in many proteolytic clostridia and Gram-positive bacteria mainly via the activity of the glycine reductase system [50]. We have previously identified, using mass spectrometry, some enzymes of these pathways in T. denticola which were abundant suggesting glycine catabolism may be a major energy source for this spirochete [27]. The importance of the glycine reductase system is also suggested by the recent report demonstrating inhibition of T. denticola growth through the impairment of selenoprotein production such as Protein A and B of the glycine reductase system by stannous salts and auranofin [51]. In support of the contention that glycine catabolism is important in T. denticola free glycine was rapidly depleted from T. denticola growth medium and the addition of glycine significantly increased T. denticola final cell density. Glycine was catabolized by T. denticola with approximately 73% of labeled [U-13 C]glycine carbon being incorporated into acetate and the majority of the remainder being incorporated into lactate, suggesting that the majority of the glycine was reduced by the glycine reductase system, where ATP is generated via substrate level phosphorylation [52,53]. The mechanism by which glycine is catabolized to lactate is less well defined but could involve conversion to serine and pyruvate, with production of NAD ( Figure 8). Hence, the preferential use of exogenously acquired glycine for catabolism rather than protein biosynthesis is consistent with glycine catabolism having a major role in energy production in T. denticola. P. gingivalis relies mainly on the uptake of peptides as a source of amino acids for energy production and it transports few free amino acids [48,49]. In mono-culture, P. gingivalis did not utilize free glycine and the concentration of free glycine in the culture supernatant increased slightly over time. The addition of cell-free T. denticola conditioned medium to a P. gingivalis culture significantly increased the concentration of free glycine in the culture medium. This stimulation of free glycine production is consistent with the increased expression of T. denticola genes encoding glycine catabolic pathways and might partially explain the increase in T. denticola cell numbers during co-culture. The increase in free glycine was dependent on the presence of P. gingivalis and a result of an increase in the hydrolysis of glycinecontaining peptides in the medium not de novo synthesis of glycine by P. gingivalis. The transcriptomic data of this current study indicated that two P. gingivalis genes (PG0753 and PG0383) encoding putative proteases were significantly up-regulated during co-culture with T. denticola. It is possible that these up-regulated enzymes acting in concert are involved in the observed increase in free glycine. Whatever the precise mechanism this result represents the first demonstration of the stimulation of peptide hydrolysis by P. gingivalis to release free glycine in response to the presence of T. denticola conditioned culture fluid. During infection of a host, P. gingivalis glycine production would be beneficial for establishment and growth of T. denticola, especially as the bacterium is motile and may respond chemotactically to the amino acid.
Transcriptome analysis of T. denticola gene expression as a result of co-culture with P. gingivalis identified the up-regulation of genes encoding several known T. denticola virulence factors including dentilisin protease complex, major sheath protein and cystalysin, which may in part explain the observed synergistic pathogenicity of T. denticola and P. gingivalis co-infections in animal models [16,18,19,[54][55][56][57][58]. As some of these proteins have also been shown to be involved in coaggregation of T. denticola with P. gingivalis [59][60][61][62] the up-regulation of these genes might enhance the coaggregation we observed in co-culture and aid site colonisation and synergistic biofilm formation by these species.
Collectively the results of this study indicate that P. gingivalis and T. denticola sense and respond to each other's presence and exhibit metabolic symbioses during co-culture which may contribute towards their establishment and persistence in the periodontal pocket. The co-aggregation, metabolite cross-feeding and upregulation of T. denticola genes encoding virulence factors help explain the temporal and spatial co-localization of the two species as surface microcolonies in subgingival plaque closely associated with chronic periodontitis [6,12] and the synergistic virulence of these bacteria in animal models of disease [18]. Figure S1 Correlation between microarray and qRT-PCR expression ratios. T. denticola mono-culture versus coculture gene expression ratios obtained using microarray or qRT-PCR were plotted and the correlation of coefficient determined by linear regression line (R 2 = 0.9839). (TIF) Figure S2 NMR spectra of the metabolic end products of glycine metabolism by T. denticola. The metabolic end products of glycine metabolism by T. denticola were determined by following the fate of [U-13 C]glycine (5 mM) added to a 24 h batch-grown T. denticola culture. Samples were collected every 24 h, filtered and the identity of the isotopically-labeled carboncontaining compounds were identified using NMR spectroscopy. Acetate and lactate were the major end products of T. denticola glycine metabolism. Protocol S1 Gas Chromatography -mass spectrometry (GC-MS).

(DOC)
Table S1 Sequence of primers used in quantitative reverse transcription PCR.
(DOC) Table S2 T. denticola genes differentially expressed during co-culture with P. gingivalis. Shading indicates genes predicted to be polycistronic. (DOC) Table S3 P. gingivalis genes differentially expressed during co-culture with T. denticola. Shading indicates genes are predicted to be polycistronic. (DOC) Figure 8. Proposed T. denticola glycine catabolic pathways. Glycine can be oxidized by the glycine cleavage system (1), producing NH 3 , CO 2 and CH 2 -THF. Glycine and CH 2 -THF can be condensed to form serine by serine hydroxymethyltransferase (2). Serine is deaminated to produce pyruvate by serine dehydratase (3). Lactate dehydrogenase (4) catalyzes the interconversion of pyruvate and lactate with concomitant interconversion of NADH and NAD + . Pyruvate can also be metabolized to acetate by pyruvate-ferredoxin oxidoreductase (5), phosphate acetyltransferase (6) and acetate kinase (7). Glycine can also be reduced to acetyl-P by the glycine reductase system (8). doi:10.1371/journal.ppat.1003955.g008