A novel phosphoglycerol serine-glycine lipodipeptide of Porphyromonas gingivalis is a TLR2 ligand.

Porphyromonas gingivalis is a Gram-negative anaerobic periodontal microorganism strongly associated with tissue-destructive processes in human periodontitis. Following oral infection with P. gingivalis, the periodontal bone loss in mice is reported to require the engagement of Toll-like receptor 2 (TLR2). Serine-glycine lipodipeptide or glycine aminolipid classes of P. gingivalis engage human and mouse TLR2, but a novel lipid class reported here is considerably more potent in engaging TLR2 and the heterodimer receptor TLR2/TLR6. The novel lipid class, termed Lipid 1256, consists of a diacylated phosphoglycerol moiety linked to a serine-glycine lipodipeptide previously termed Lipid 654. Lipid 1256 is approximately 50-fold more potent in engaging TLR2 than the previously reported serine-glycine lipid classes. Lipid 1256 also stimulates cytokine secretory responses from peripheral blood monocytes and is recovered in selected oral and intestinal Bacteroidetes organisms. Therefore, these findings suggest that Lipid 1256 may be a microbial TLR2 ligand relevant to chronic periodontitis in humans.

actions and tissue destruction, the virulence factors responsible specifically for the tissue-destructive processes in periodontitis are only partially characterized.
P. gingivalis is known to engage the innate immune receptor, Toll-like receptor 2 (TLR2), when promoting inflammatory and tissue destructive processes associated with experimentally-produced periodontitis in animals (3)(4)(5). Several virulence factors of P. gingivalis have been reported to engage TLR2, including lipopolysaccharide (LPS), fimbriae, and others (6)(7)(8)(9)(10). More recent reports have discounted LPS as a TLR2 ligand by either purifying LPS or its Lipid A and demonstrating essentially no TLR2 activity (11) or by synthesizing Lipid A of P. gingivalis and demonstrating no TLR2 activity (12). Another report has identified a lipoprotein lipase-sensitive TLR2 ligand that is a contaminant of LPS recovered from P. gingivalis, and although the product was not chemically characterized, it likely accounts for the TLR2 engagement previously attributed to LPS (13). Serine-glycine lipodipeptides of P. gingivalis, including Lipid 654 and Lipid 430 (14), and glycine aminolipids of P. gingivalis, including Lipid 567 and Lipid 342 (15), are TLR2 ligands, but these lipids are not as potent on a mass basis as the common lipopeptide TLR2 ligands Pam2Cys and Pam3Cys (15). These serine/glycine lipids represent approximately 5% of the total lipids recovered from P. gingivalis (14,15) and could account for at least a portion of the TLR2 engagement observed with the total lipid extract of P. gingivalis. However, evidence presented here indicates that a new lipid class of P. gingivalis is between 10 and 100 times more potent than the serine/ glycine lipids of P. gingivalis in engaging TLR2. Structural characterization revealed that this novel lipid class is a diacylated phosphoglycerol serine-glycine lipodipeptide, and it was recovered in all Bacteroidetes organisms tested, including oral and intestinal isolates. The current investigation also demonstrates the capacity of this new lipid class to promote cytokine secretory responses in human peripheral blood monocytes in culture.
Abstract Porphyromonas gingivalis is a Gram-negative anaerobic periodontal microorganism strongly associated with tissue-destructive processes in human periodontitis. Following oral infection with P. gingivalis, the periodontal bone loss in mice is reported to require the engagement of Toll-like receptor 2 (TLR2). Serine-glycine lipodipeptide or glycine aminolipid classes of P. gingivalis engage human and mouse TLR2, but a novel lipid class reported here is considerably more potent in engaging TLR2 and the heterodimer receptor TLR2/TLR6. The novel lipid class, termed Lipid 1256, consists of a diacylated phosphoglycerol moiety linked to a serine-glycine lipodipeptide previously termed Lipid 654. Lipid 1256 is approximately 50-fold more potent in engaging TLR2 than the previously reported serine-glycine lipid classes. Lipid 1256 also stimulates cytokine secretory responses from peripheral blood monocytes and is recovered in selected oral and intestinal Bacteroidetes organisms. Therefore, these findings suggest that Lipid 1256 may be a microbial TLR2 ligand relevant to chronic periodontitis in humans.
Supplementary key words phospholipids • lipid signaling • tandem mass spectrometry • monocytes • high-performance liquid chromatography • periodontitis • Bacteroidetes Porphyromonas gingivalis is considered to be a primary periodontal pathogen in promoting chronic adult periodontitis, a chronic inflammatory condition that is associated with loss of attachment and bone around teeth (1). This organism produces a variety of potential virulence factors that are common cell-membrane constituents recognized by the innate immune system (2). These virulence factors are thought to gain entry to the gingival tissues surrounding teeth, and, once present in tissues, they elicit a chronic inflammatory response that is associated with attachment loss around teeth and bone breakdown. While there is considerable evidence supporting the concept that virulence factors are the primary cause of inflammatory re-

Reagents and cell lines
P. gingivalis strain 33277 and other oral Bacteroidetes bacteria were obtained from ATCC. Intestinal Bacteroidetes species were a gift from Sydney Finegold. Bis(trimethylsilyl)trifluoroacetamide (BSTFA) was obtained from Sigma-Aldrich. Broth media for bacterial culture was obtained from Gibco. HPLC columns were obtained from Sigma-Aldrich, and HPLC solvents and other organic solvents were obtained from Thermo Fisher Scientific. Antihuman antibody preparations and Pam2Cys and Pam3Cys were obtained from InvivoGen.

Bacterial lipid extraction
P. gingivalis and all other bacteria were grown in broth medium containing basal [peptone, trypticase, brain heart infusion, and yeast extract (BBL; Thermo Fisher Scientific)] medium supplemented with hemin and menadione (Sigma-Aldrich) (16). The suspension cultures were incubated in an anaerobic chamber flushed with N 2 (80%), CO 2 (10%), and H 2 (10%) at 37°C for 4-5 days, and the bacteria were harvested by centrifugation at 3,000 g for 2 h. Samples of bacterial cultures were examined following the Gram stain. After lyophilization, a P. gingivalis pellet (4 × 2 g) was extracted overnight using the phospholipid extraction procedure of Bligh and Dyer (17) as modified by Garbus et al. (18). Specifically, 4 ml H 2 O + 16 ml MeOH:CHCl 3 (2:1 v/v) were added to each bacterial sample (2 g) and vortexed. After 12 h, 6 ml 2 N KCl + 0.5 M K 2 HPO 4 and 6 ml CHCl 3 were added, and the sample was vortexed. After centrifugation at 2,500 g for 2 h, the lower organic phase was carefully removed, and 6 ml CHCl 3 was added to each sample and vortexed. After centrifugation, the CHCl 3 phase was removed and combined with the previous organic solvent extract. The lipid extracts from P. gingivalis were dried under nitrogen.

HPLC methods
Fractionation of bacterial lipids by HPLC was accomplished using a semipreparative HPLC column (1 × 25 cm silica gel; Ascentis, 5 m; Sigma-Aldrich) and eluting lipids isocratically with hexane-isopropanol-water (6:8:0.75; v/v/v) (19). Lipid samples were dissolved in HPLC solvent to achieve a concentration of approximately 14 mg/ml. Each sample was centrifuged at 2,500 g for 10 min, and the supernatant was removed for HPLC fractionation. Semipreparative HPLC fractionation was accomplished using an HPLC system equipped with dual pumps (LC-10ADvp), an automated controller (SCL-10Avp), and an in-line UV detector (SPD-10Avp; Shimadzu Scientific Instruments). For each chromatographic separation, 7-10 mg of lipid was applied, and fractions were pooled for 80 column fractionations. Samples were eluted at 1.4 ml/min and monitored at 205 nm, with 2 min fractions collected over a total of 2 h. Fractions were dried under nitrogen and resuspended in CHCl 3 . Lipid recovery in each HPLC fraction was determined by drying 5 l from each fraction and weighing the sample using a Cahn Electrobalance.

Bacterial fatty acid analysis
Total fatty acids within specific pooled acidic fractions (HPLC fractions 17-24; see Results) were first analyzed by treating lipid aliquots (10 g) with 6N HCl (0.2 ml, microwaved for 90 s). The hydrolysates were cooled and dried under nitrogen. The samples were treated with BSTFA overnight at room temperature to generate TMS derivatives of fatty acids and other products. Replicate samples were transferred to autosampler vials for GC-MS analysis. Synthetic standards of isobranched fatty acids, saturated fatty acids, and 3-OH fatty acids were obtained from Matreya, Inc. and treated to form TMS derivatives using the same method.
Fresh sodium methoxide (NaOCH 3 ) was prepared by dissolving metallic sodium in dry methanol. Ester-linked fatty acids were then determined by treating aliquots of the pooled acidic fractions with fresh NaOCH 3 (0.5 N × 0.5 ml for 30 min at 50°C). At the completion of the hydrolysis reaction, the sample was acidified with 1N HCl, and fatty acid methyl esters were extracted into chloroform (2 ml × 3 extractions) and dried under nitrogen. The fatty acid methyl esters were then treated with BSTFA overnight at room temperature, and samples were then analyzed by GC-MS as described below.

Bacterial serine analysis
Additional aliquots of the pooled acidic fractions (HPLC fractions 17-24; see Results) were treated to form methyl ester-pentafluoropropyl ether/amide derivatives for analysis according to the method of Fuchs et al. (21). The dried samples, including a serine synthetic standard, were first treated with acetyl chloride-methanol (1:4; v/v; 100 l; 70°C for 45 min) and dried. The samples were then treated with chloroform-pentofluoropropionic anhydride (4:1; v/v; 500 l; 100°C for 20 min) and dried. The residues were dissolved in chloroform and analyzed by negative chemical ionization GC-MS.

GC-MS analysis
GC-MS was carried out on a Hewlett Packard 5975C gas chromatograph mass spectrometer. Fatty acid samples were applied to an HP-5MS column (30 m × 0.25 mm, 0.25 m film thickness; Agilent Technologies) held at 100°C. Fatty acid derivatives were eluted using a temperature program of 20°C/min from 100°C to 290°C (held for 5 min). Products were eluted using a continuously adjusted helium pressure at the column head to maintain constant flow of helium during the column heating. The injector block was held at 285°C, and the transfer tube was maintained at 290°C. Mass spectra were acquired using combined electron impact and selected ion detection modes. Bacterial fatty acid products were identified by the retention time of characteristic positive ions. Fatty acid identities were verified using synthetic fatty acid standards. The serine derivatives (pentafluoropropyl ether/amide-methyl ester derivatives) in each sample were eluted using the same temperature program and column as described above, and mass spectra were acquired using the negative chemical ionization detection mode.

LC-MS analysis
The analysis of lipids within individual HPLC fractions and in pooled fractions utilized the mass spectrometric instrumentation located at the University of Connecticut Center for Environmental Sciences and Engineering. Routine analysis of lipid samples used a Waters Acquity UPLC coupled with an Acquity TQD tandem mass spectrometer. An Acquity UPLC CSH C18 (1.7 µm, 2.1 × 100 mm) column, heated to 50°C and with a sample injection volume of 20 µl on a 20 µl loop, was utilized for analyte separation. The mobile phase consisted of 10 mM ammonium formate [0.1% formic acid in 40% water/60% acetonitrile (solvent A)], and 10 mM ammonium formate [0.1% formic acid in 90% isopropyl alcohol/10% acetonitrile (solvent B)] was used for gradient elution. An initial flow of 75% solvent A was held for 3 min before increasing linearly to 100% solvent B until 8 min, after which the column was reconditioned to the initial state for another 1 min. The total run time was 9 min, with a constant flow rate of 0.4 ml/ min.
The detection and quantification of analytes and surrogate internal standard was performed in negative ESI-MS/MS mode (multiple-reaction monitoring) using Waters IntelliStart software for analyte signal optimization. Statistical analysis for obtaining calibration and quantification results for all compounds was performed using Waters QuanLynx, which is included in MassLynx version 4.2. Parameters for the mass spectrometer were set as follows: capillary voltage, 2.0 kV; cone voltage, 30 V; desolvation temperature, 400°C; source temperature, 120°C; desolvation gas flow, 750 l/h; and collision gas flow, 0.2 ml/min.
Pooled lipid preparations were analyzed using a Synapt G2-Si HDMS (qTOF) instrument (Waters). A sample of pooled lipid was dissolved in methanol at a concentration of approximately 1 mg/ml and was manually injected at 30 µl/min. Mass spectra were acquired using MS and MS/MS formats in both negative and positive ionization modes. Positive-mode MS analysis implemented the following parameters: 500-1500 m/z mass range, 0.1 s scan time, 2 kV capillary voltage, 100°C source temperature, 30 V sample cone, 200°C desolvation temperature, 800 l/h desolvation gas flow, and targeted MS/MS scans that used a collision-induced dissociation (CID) energy of 50 V. Negative-mode MS analysis implemented the identical parameters as above except for the following: 1.8 kV capillary voltage, 20 V sample cone, and targeted MS/MS scans that used CID fixed energies ranging from 40 to 50 V.

NMR analysis
All NMR experiments were performed on a Varian 800 MHz NMR spectrometer equipped with an HCN room temperature probe except the 1D 31 P experiment, which was performed on a Varian 400 MHz instrument equipped with a BB probe tuned to C HMBC spectra were collected using standard pulse sequences. The 1D 1 H spectrum was collected with 512 transients, a 5 s recycle delay, 16k complex points, and a bandwidth of 12,019.23 Hz centered at 4.078 ppm. The COSY was collected with 8 transients, a 1 s recycle delay, 4k (t 2 ) × 1k (t 1 ) complex points, and a bandwidth of 6,983.24 Hz centered at 4.078 ppm in both dimensions. The HSQC was collected with 64 transients, a 1 s recycle delay, 2k (t 2 ) × 300 (t 1 ) complex points, and bandwidths of 6,983.24 Hz (t 2 ) and 39,396 Hz (t 1 ) centered at 4.078 ppm (t 2 ) and 93.08 ppm (t 1 ). The HMBC was collected with 128 transients, a 1.8 s recycle delay, 4k (t 2 ) × 400 (t 1 ) complex points, and bandwidths of 6,983.24 Hz (t 2 ) and 44,198.90 Hz (t 1 ) centered at 4.078 ppm (t 2 ) and 105.59 ppm (t 1 ) optimized for 8 Hz 2-3-bond couplings with 1 H-13 C 1-bond suppression. The HMBC was collected in magnitude mode but with the phase arrayed, resulting in two spectra that were processed separately and coadded, effectively doubling the transients. The 1D 31 P spectrum was acquired with 16,384 transients, a 8 s recycle delay, 128k complex points, and a bandwidth of 70,000 Hz centered at 31.92 ppm. Data were processed with NMRPipe (22) and analyzed with MNova from Mestrelab Research, both provided in the NMRbox platform (23). The program ALATIS was used for naming the Lipid 1256 class (24).

Lipid preparation for cell cultures
Predetermined amounts of the pooled acidic fractions (HPLC fractions 17-24; see Results) and other lipid preparations were transferred to conical glass vials and dried under nitrogen. Each lipid fraction was then supplemented with an aliquot of cell culture medium. The lipids were sonicated for 15 s at 3 W, after which the stock lipid preparations were added to cells in culture to achieve the indicated doses.

HEK cell culture
HEK cells (InvivoGen) were used to evaluate human Toll-like receptor engagement by P. gingivalis lipids. According to the commercial supplier, endogenous genes for TLR1 or TLR6 were neutralized by double knockout before transfection with either human TLR2/6 or TLR2/1. Both cell lines were also transfected with human CD14. The engagement of the indicated Toll-like receptors leads to the activation and translocation of NF-B together with the expression of the secretory alkaline phosphatase (SEAP) reporter gene and subsequent release of SEAP from the HEK cells. SEAP is then detected in culture medium using the development medium provided by InvivoGen. TLR2, TLR4, TLR2/6, and TLR2/1 transfected HEK cell lines as well as Null 1 HEK cells were grown in DMEM medium supplemented with antibiotics, antimycotics, and 10% fetal calf serum as recommended by the commercial supplier. Sonicated lipid preparations as well as suitable positive control preparations (Pam2Cys and Pam3Cys) and Salmonella typhimurium LPS (Sigma-Aldrich) were added to 96-well culture dishes to achieve the indicated concentrations. Controls included the treatment of HEK cells with anti-human TLR2 or TLR6 neutralizing antibodies versus suitable isotype antibody controls (InvivoGen). Antibody preparations were added to achieve a final concentration of 10 g/ml. Wells were then supplemented with HEK-Blue Detection medium containing the HEK cells and incubated for 22-26 h. The plates were read at 630 nm for quantification SEAP activity.

Peripheral blood monocyte cell culture
Fresh peripheral blood was obtained from consenting volunteers in compliance with institutional review board-approved protocol. Blood (50 ml) obtained by venipuncture was processed for negative selection of monocytes using the StemCell Easy Sep monocyte purification kit. The highly enriched monocytes were counted electronically and plated at 130,000 to 150,000 cells per well in 24-well culture dishes. RPMI 1640 medium with 10% fetal calf serum, 5 mM glutamine, 100 U/ml penicillin and 100 g/ml streptomycin was used for cell culture. After 1 h for cell adherence, monocytes were treated with the indicated bacterial lipids or synthetic lipopeptide preparations (Pam2Cys). After a 24 h incubation, the medium was harvested and frozen. At the completion of lipid treatment experiments, specific culture wells were routinely stained with FITC-anti-human CD14 versus isotype control (both from InvivoGen) for verification of monocyte purity (>95%).

ELISA for TNF- and IL-1
Previously isolated media samples were thawed and analyzed for TNF- and IL-1 without dilution. TNF- or IL-1 was quantified using commercially available ELISA kits (R&D Systems).

Statistical analysis
Statistical tests included the calculation of sample means and standard deviations. Histogram bars depict the sample means and error bars represent standard deviations (the number of samples analyzed are listed in the figure legends). One-factor ANOVA with pairwise comparisons using Dunnett's multiple comparisons test was then used to determine significant differences between treatment group.

Isolation of the novel lipid class
Total lipids of P. gingivalis were fractionated using a neutral solvent semipreparative HPLC fractionation (Fig. 1A) as previously described (16). In prior work, fractions were typically collected only for approximately 60 min and lateeluting minor lipid constituents were discarded. While the serine-glycine lipodipeptide lipids and the glycine aminolipids were interspersed with phospholipids and complex lipids as they emerged during the first h (14,15), two late eluting lipid peaks were identified when the fractionation was extended to 2 h. For the present study, the larger peak emerging between 94 and 95 min was evaluated. By combining approximately 80 fractionations from two independent batches of total lipids of P. gingivalis, approximately 6 mg of lipid was recovered in fractions eluting between 90 and 100 min (Fig. 1A). Fractions eluting from 90 to 100 min demonstrated significant activation of human TLR2 in a human embryonic kidney (HEK reporter cells) cell screen (see method below) and were pooled.
The pooled fractions were then eluted over the same semipreparative column using solvent A supplemented with 0.1% acetic acid (Fig. 1B) (20). The lipids emerged with a much shorter retention times, suggesting protonation of at least one acidic moiety within these fractions. Aliquots were taken from each fraction and were evaluated for engagement of human TLR2 again using transfected HEK reporter cells. A peak of TLR2 engaging activity was observed with fractions 17-24 min with strong responses observed at doses as low as 10 ng/ml (Fig. 1C). Fractions 17-24 were pooled and evaluated for lipid products using LC-MS and MS/MS approaches as described below.

Structural characterization of the novel lipid
The pooled 17-24 min fractions were evaluated for MS characteristics, including negative ion constituents using a Synapt G2-Si instrument.  (Table 1). MS/MS of these negative parent ions recovered ions consistent with saturated fatty acids ranging from 13 to 18 carbons in length. MS/MS of the m/z 1255.9459 ion revealed a dominant fatty acid negative ion (m/z 241.2196) consistent with the presence of C 15:0 . A sample of the Lipid 1256 preparation was hydrolyzed in 6N HCl, and the hydrolysis products were treated to form TMS derivatives followed by electron-impact GC-MS analysis. Selective ion monitoring revealed that iso and anteiso C 15:0 are the dominant fatty acids in the Lipid 1256 preparation with lesser amounts of 3-OH iso C 17:0 and C 16:0 and C 18:0 present in this lipid class ( Fig. 2A). The relative levels of total fatty acids recovered indicate that C 15:0 fatty acid is approximately three times more prevalent than 3-OH iso C 17:0 . Lesser amounts of C 13:0 , C 16:0 , and C 18:0 were also recovered. Sodium methoxide treatment of Lipid 1256 revealed methyl esters of iso + anteiso C 15:0 and a substantially lower amount of C 16:0 recovered, indicating that these fatty acids are held in ester linkage (Fig.  2B). Serine was also detected as a methyl ester, pentafluoropropyl derivative in the hydrolysis products (Fig. 2C,  D). The unknown peak shown in Fig. 2A yielded the mass spectrum shown in Fig. 1E, which included a TMS ion fragment (m/z 73 ion). The molecular ion of the unknown  Table 1). The m/z 152.9941 fragment is consistent with a deacylated and dehydrated phosphoglycerol constitutent (Fig. 3D). From these results, we conclude that the new lipid class contains phosphoglycerol, serine, and the fatty acids listed above, and the dominant ester-linked fatty acids include iso and anteiso C 15:0 and a relatively low amount of C 16:0 . The remaining fatty acids, particularly 3-OH iso C 17:0 , were recovered only in the acid-hydrolyzed products of Lipid 1256 and are presumed to be amide-linked.
The 1 H and 13 C signals observed with NMR evaluation of Lipid 1256 are listed in Table 2. Within the COSY and HSQC NMR spectra, 1 H and 13 C signals for carbons 10-25 are consistent with previously reported NMR spectra for the serine-glycine lipodipeptide lipid, Lipid 654 (14), including serine, glycine, amide-linked 3-OH iso C 17:0 , and ester-linked iso C 15:0 . However, the 1 H and 13 C signals for serine were shifted slightly downfield from those reported previously for Lipid 654 (14). An additional three-carbon spin system was identified in Lipid 1256 that is consistent with the presence of a diacyl substituted glycerol. Carbons 7, 8, and 9 represent the three-carbon spin system determined by COSY with two protons on carbons 7 and 9 and one proton on carbon 8. NMR also revealed the presence of 31 P in the L1256 structure. Correlations with the HMBC are supportive of the structure except for four missing correlations. Integration of minor proton signals indicate possible contamination with two different products, one contaminant representing about 14% and the other product representing about 5% of the normalized signal for the single proton on carbon 11 (Table 2). Figure 3A shows the proposed structure of the dominant lipid homolog of Lipid 1256. This structural reconciliation shows that the Lipid 1256 class represents Lipid 654, linked to a diacylated glycerol phosphate. MS/MS analysis of the dominant homolog within the Lipid 1256 class revealed negative ion fragments (Table 1) consistent with the proposed negative ion fragments depicted in Fig. 3B-D. By extension, the other species of the Lipid 1256 class would be formed by substituting other fatty acid constituents within the respective ion fragments. Based on the fragmentation of the m/z 1255.9459 parent ion, the phosphate of the phosphoglycerol moiety must be linked to the serine hydroxyl because no other structural variants will generate ion fragments with the observed masses. The phosphate constituent therefore acts as a linker between the Lipid 654 core structure and the diacylated glycerol moiety. Altogether, this evidence demonstrates that the Lipid 1256 class is a phosphoglycerol serine-glycine lipodipeptide, a new lipid class of P. gingivalis.

TLR2 engagement by the novel lipid
The Lipid 1256 preparation was then tested for engagement of TLR2 and related receptors using human embryonic kidney (HEK) NF-B SEAP reporter cells transfected with specific human TLRs (Fig. 4). Cumulative SEAP release into culture medium was measured by substrate conversion measured at 630 nm (Fig. 4A). Comparing dose responses from Lipid 1256 with canonical TLR2 ligands, including Pam2Cys (TLR2/6 ligand) and Pam3Cys (TLR2/1 ligand) (26), revealed that Lipid 1256 is about 10-fold less potent on a mass basis (5-fold less on a mole basis) than Pam2Cys in engaging TLR2 or TLR2/6 (Fig. 4A). Lipid 1256 and Pam2Cys did not engage TLR2/1, but Pam3Cys engaged TLR2/1 as expected (26). Lipid 1256 did not activate the nontransfected Null 1 HEK cells or HEK cells transfected with TLR4 (receptor for LPS). Furthermore, Lipid 1256 responses were blocked by treatment with anti-human TLR2 neutralizing antibody and anti-human TLR6 neutralizing antibody, but the isotype antibody controls did not interfere with Lipid 1256 activation of TLR2 or TLR2/6 transfected HEK cells. Finally, Lipid 1256 did not stimulate TLR2/TLR1 transfected HEK cells, indicating that Lipid 1256 is primarily a TLR2/TLR6 ligand. A sample of enriched Lipid 1256 was dissolved in methanol at a concentration of 1 mg/ml and was manually infused into a Synapt G2-Si mass spectrometer. Mass spectra were first acquired for the unfragmented parent ions followed by MS/MS evaluation of collision-induced fragments from each parent negative ion observed in the Lipid 1256 preparation. The ions shown in the table were observed using a collision energy of 50 V. The mass axis calibration was locked. *Dominant parent negative ion homolog within the Lipid 1256 class. ^Dominant CIDnegative ions from the indicated parent negative ions.
As we have previously reported, Lipid 654 and Lipid 567 activate HEK cells through the engagement of either TLR2 or TLR2/TLR6 but not TLR2/TLR1 (15). Dose-response characteristics of Lipid 1256 were compared with Lipid 654 and Lipid 567 (Fig. 4B). Lipid 1256 (10 ng/ml) activated HEK TLR2 and TLR2/6 cells to approximately the same degree as 1 g/ml of either Lipid 654 or Lipid 567 lipid preparations that were previously isolated from P. gingivalis (14,15). This comparison revealed that Lipid 1256 is between 10 and 100 times more potent than either Lipid 654 or Lipid 567 in activating human TLR2 or TLR2/TLR6 (Fig. 5B).
Anti-human TLR2, TLR6, and TLR1 neutralizing antibodies were only partially effective in attenuating Pam2Cys and Pam3Cys engagement at the doses used in Fig. 4A. The effect of decreasing the dose of either Pam2Cys or Pam-3Cys was evaluated next. Decreasing doses of these ligands were combined with either anti-human TLR2, TLR6, or TLR1 neutralizing antibodies versus isotype controls (Fig.  4C) and tested in the listed HEK cell types. Figure 4C shows that the effect of Pam2Cys at a dose of 0.1 ng/ml is virtually completely attenuated by anti-human TLR2 neutralizing antibody in HEK TLR2 cells as well as in HEK TLR2/TLR6 Fig. 2. Fatty acids, serine, and phosphoglycerol constituents identified from hydrolyzed Lipid 1256 of P. gingivalis. Lipid 1256 was hydrolyzed in 6N HCl (0.2 ml, microwaved for 90 s). The contents were dried and treated with BSTFA overnight to form TMS derivatives. The identity of fatty acids released from L1256 was confirmed with known standards (A). The unknown product was further evaluated as shown in panels E and F. Aliquots of Lipid 1256 were hydrolyzed in sodium methoxide (2N, 0.5 ml, 50°C, 30 min), and the lipids were extracted into chloroform. The dried fatty acid methyl esters were then treated overnight with BSTFA at room temperature and evaluated by GC-MS (B). Serine standard was treated to form a methyl ester, pentafluoropropyl derivative and was evaluated by negative chemical ionization GC-MS (C). Hydrolyzed products from Lipid 1256 were treated to form methyl ester, pentafluoropropyl derivatives and were evaluated by negative chemical ionization GC-MS (D). Mass spectrum (E) of the unknown product shown in panel A and structural reconciliation of unknown product (F). cells by anti-human TLR6 neutralizing antibody. By contrast, Pam3Cys at a dose of 0.01 ng/ml was largely attenuated only by anti-human TLR1 antibody in HEK TLR2/ TLR1 cells. Anti-human TLR2 neutralizing antibody inhibition of Pam3Cys effects in HEK TLR2 cells was observed primarily at a dose of 0.001 ng/ml. Therefore, the neutralizing antibody blocking of Pam2Cys and Pam3Cys effects in the respective transfected HEK cells occurs at relatively low doses of these ligands. The specificity of Pam2Cys and Pam-3Cys effects on TLR2/TLR6 and TLR2/TLR1, respectively, is confirmed with this survey. Figure 4C also shows that the neutralizing antibodies or their respective isotype controls do not stimulate SEAP release from transfected HEK cells or Null HEK cells.

Peripheral blood monocyte responses
For the evaluation of biological responses in primary cultures of cells, human peripheral blood monocytes were isolated from fresh blood of consenting donors and were placed into culture (three independent blood donations). Following a 24 h treatment with the indicated agents, me-dia samples were collected and assayed for both IL-1 and TNF- by ELISA (Fig. 5). TNF- and IL-1 release was significantly elevated for freshly isolated monocytes when treated for 24 h with either Lipid 1256 (1.0 g/ml) or Pam-2Cys (0.1 g/ml). Reducing the Lipid 1256 dose to 0.1 g/ ml resulted in a loss of statistical significance for both cytokine secretory responses, and Lipid 654 at 1 g/ml did not promote significant cytokine secretory responses. These results indicate that Lipid 1256 will promote significant cytokine secretory responses in primary cultures of peripheral blood monocytes.

Lipid 1256 recovery in bacterial lipid isolates
The recovery of Lipid 1256 relative to other serine-glycine lipids in P. gingivalis lipid extracts as well as the recovery of Lipid 1256 relative to other serine-glycine lipids in total lipid extracts from other oral Bacteroidetes species and two intestinal Bacteroidetes was investigated. Figure 6 shows that Lipid 1256 and Lipid 567 are the dominant serine-glycine lipids recovered from Bacteroidetes species, including multiple isolates from the type strain of P. gingivalis.  Table 1. Sample preparation and instrument conditions used for data acquisition are listed in Table 1 Table 1 shows that the fragment shown in panel D is detected in all five species of Lipid 1256. Some variability is evident in the recovery of Lipid 1256 between total lipid isolates from P. gingivalis, but all of the species and isolates tested contain Lipid 1256. It is likely that Lipid 1256 is a complex lipid class consistently recovered in all members of the Bacteroidetes phylum.

DISCUSSION
The periodontal pathogen P. gingivalis promotes important host-cell responses through the engagement of TLR2. Contrasted with the serine/glycine lipids of P. gingivalis (Lipids 342, 430, 567, and 654), the current article demonstrates that a new lipid product of this organism, Lipid 1256, represents a modified serine-glycine lipodipeptide that is a relatively strong ligand for TLR2. Recent work has shown that serine-glycine lipodipeptide lipids (Lipid 654 and Lipid 430) and glycine aminolipids (Lipid 567 and Lipid 342) of P. gingivalis engage both human and mouse TLR2, and these lipids, although they are not potent TLR2 ligands compared with canonical TLR2 ligands such as Pam2Cys, can mediate human TLR2 responses when prepared as highly enriched bacterial lipid preparations or as Refer to Fig. 3A for carbon assignments. All integrals were normalized against the single proton peak for carbon 11. All expected correlations in the DQF COSY were observed, and all but four correlations in the HMBC were observed and are indicated with an asterisk. A 1D 31 P NMR spectrum showed the presence of a single 31 P peak at 0.98 ppm. a Proton peak groupings corresponding to carbons 1,2; 4,5; and 21,22 are very close in both 1 H and 13 C chemical shifts, and it is not possible to determine which of the three fatty acids the pairs belong to. b Proton peaks for carbons 1, 4, 18, and 22 overlap and integrate together to 8.84, and each was assigned an integral of 2.21.
c Proton peaks for carbons 2, 5, and 21 overlap and integrate together to 6.9, and each was assigned an integral of 2.3. d Proton peaks for carbons 7, 12, 14, and 16 correspond to CH 2 groups and have two distinct chemical shifts demonstrating restriction from free rotation. For DQF COSY and HMBC correlations, a  denotes a second proton peak for the given carbon. e The proton peak for carbon 7 at 4.41 ppm integrates to 1.09, while the peak at 4.18 ppm overlaps and cannot be integrated; thus, the integral was taken as 2(1.09) = 2.18. f Proton peaks for carbons 8 and 17 overlap and integrate together to 2.0 and each is assigned an integral of 1.00. g Proton peaks for carbons 9 and 14 overlap and integrate together to 5.37 and each was assigned an integral of 2.69. Overlapping contaminants cause the area to integrate higher than expected.
h The proton peak for carbon 12 at 4.26 ppm integrates to 1.05, while the peak at 4.16 ppm overlaps and cannot be integrated; thus, the integral was taken as 2(1.05) = 2.10.
i Cannot determine integral because the proton peak overlaps with the large signal composed of the CH 2 groups of the four fatty acids.
j Proton peaks corresponding to carbons 23, 24, and 25 for the four fatty acids show perfectly overlapping signals demonstrating that the branched methyls are identical in all four fatty acids. Integrals for peaks 23, 24, and 25 were divided by 4, 4, and 8, respectively. k The proton peak corresponding to the overlapping CH 2 groups in the fatty acids correspond to 33 carbons and 66 protons. In addition, the proton peak for carbon 19 also overlaps. The integral was thus divided by 34. synthetic homologs (14,15). More importantly, these lipids are related to each other in that all contain a core structure of 3-OH iso C 17:0 amide linked to glycine, which by itself was termed Lipid 342 based on its negative ion mass. The next higher mass glycine lipid class, Lipid 567, represents the addition of an esterified fatty acid (usually iso or anteiso C 15:0 ) to Lipid 342, and Lipid 654 represents Lipid 567 with the addition of a terminal serine that is amidelinked to the glycine moiety. Lipid 430 represents Lipid 342 with serine attached to glycine. The current investigation shows that P. gingivalis can apparently modify Lipid 654 by attaching a diacylated phosphoglycerol to the serine of Lipid 654, resulting in the formation of Lipid 1256. This lipid class forms facile negative ions and is therefore presumed to be acidic, consistent with faster elution from a normal phase column when using HPLC solvent containing 0.1% acetic acid. The mass of the parent species of this lipid class also were not altered with catalytic hydrogenation using platinum catalyst and H 2 gas treatment overnight, indicating that unsaturated bonds do not exist (data not shown).
The synthetic processes in P. gingivalis and other members of the Bacteroidetes phylum responsible for the production of Lipid 1256 have not been defined but could be related to the synthesis or metabolism of other complex lipids previously identified in P. gingivalis. P. gingivalis, and other members of the Bacteroidetes phylum synthesize two glycerol phosphate dihydroceramide lipid classes (16).
However, the glycerol phosphate moiety is not acylated and if released through enzymatic hydrolysis, it could be acylated before attachment to Lipid 654 through the action of glycerol-3-phosphate acyltransferase (27). An alternative synthetic pathway could involve the phosphorylation of bacterial diacylglycerol by diacylglycerol kinase (28). Both pathways will produce phosphatidic acid, which can be activated to CDP-diacylglycerol followed by linkage of diacylglycerol phosphate as a phosphoester to the serine of Lipid 654 (27,28). A more likely synthetic pathway would involve phospholipase D-mediated release of phosphatidic acid from phosphatidylethanolamine lipids previously identified in P. gingivalis (29) followed by the formation of a phosphoester to the serine of Lipid 654 (28). The glycerol moiety of P. gingivalis phosphatidylethanolamine is substituted with either iso C 15:0 (predominant form), iso C 13:0 , or lower amounts of C 16:0 or C 14:0 (16). If phosphatidylethanolamine is the primary source of diacylated glycerol phosphate for the synthesis of the Lipid 1256 lipid class, combinations of these fatty acids can account for the various parent ion species of Lipid 1256 identified in Table  1. Regarding the remaining esterified fatty acids in Lipid 1256, the  hydroxyl of 3-OH iso C 17:0 can be substituted with iso C 15:0 , C 14:0 , or iso C 13:0 based on the reported masses of the parent species within the Lipid 654 class and the esterified fatty acids recovered from this lipid class (14). The predominant fatty acid recovered in either Lipid 654 or Lipid 1256 is iso C 15:0 . Regardless of the fatty acid substitutions, the glycerol phosphate substrate attached to Lipid 654 species can be derived from several potential complex Lipid effects on cytokine secretory responses in human peripheral blood monocytes. Bacterial lipid effects on IL-1 and TNF- release from human peripheral blood monocytes were averaged for three independent blood donations. Peripheral blood monocytes were isolated as described in the Materials and Methods, and two culture wells were treated with Lipid 1256 (L1256), Lipid 654 (L654), or Pam2Cys (P2C) at the indicated doses for 24 h of culture. Each medium sample was assayed in duplicate for IL-1 and TNF- by ELISA. Histogram bars represent the treatment means ± SDs. * ,# Significant differences from control cultures (onefactor ANOVA with Dunnet's multiple comparisons tests). lipid sources previously identified in P. gingivalis. Finally, it is theoretically possible that two phosphatidic acid molecules could be linked to form cardiolipin (27,28). However, negative ions consistent with cardiolipin substituted with these fatty acids have not been identified in total lipid extracts from P. gingivalis. Future work will be directed toward understanding the synthetic pathways, including the enzymes responsible for the synthesis of Lipid 1256 as well as understanding how other fatty acids can be substituted into Lipid 1256 such as esterification and deesterification reactions from other bacterial complex lipids and perhaps host complex lipids.
Lipid 1256 is distinguished from the previously named serine-glycine lipodipeptides and glycine aminolipid classes in that it contains four fatty acids, two of which are ester-linked to a phosphoglycerol moiety. By contrast, the canonical lipopeptide Pam2Cys, which engages TLR2/ TLR6, contains two acyl chains, and the lipopeptide Pam-3Cys, which engages TLR2/TLR1, contains three acyl chains. Lipid 1256 contains four fatty acids, yet it specifically engages TLR2/TLR6 similar to a diacylated lipopeptide Pam2Cys. Previously, a lipoprotein lipase susceptible contaminant of LPS isolated from P. gingivalis was reported to engage TLR2 and TLR2/TLR1 (13). Lipid 1256 does not engage TLR2/TLR1 and therefore cannot be this contaminant. An analysis of another TLR2-active lipopeptide isolated from P. gingivalis (12) revealed a mass spectrum distinct from Lipid 1256. Therefore, we conclude that the Lipid 1256 class represents a new lipid product of P. gingivalis.
Lipid 1256-mediated TNF- and IL-1 secretory responses from freshly isolated human monocytes reveals a potency about 10-fold less than Pam2Cys. These effects suggest that this lipid class could promote a strong monocyte activation if it is deposited in host tissues associated with chronic inflammatory responses, such as with periodontal disease. Preliminary results indicate that lipid extracts from gingival tissue specimens recovered from chronic periodontitis sites contain higher median levels of Lipid 1256 compared with lipid extracts from healthy gingival tissue samples (data not shown). Therefore, the capacity of Lipid 1256 to activate monocytes is relevant to the pathogenesis of chronic periodontitis. Contrasted with monocyte responses, prior work has shown that the serine-glycine lipodipeptides Lipid 654 and Lipid 430 inhibit osteoblast differentiation in culture and markedly inhibit osteoblast gene expression with doses as low as 25 ng/ml (20). These effects were noted in osteoblasts isolated from wild-type mice, whereas osteoblasts from TLR2 knockout mice were not affected, indicating that the inhibition of osteoblast function and differentiation by these lipids is mediated through TLR2. Lipid 1256 would be expected to inhibit osteoblast function at considerably lower doses, again through the engagement of TLR2. Future experiments will evaluate osteoblast responses to Lipid 1256 in cells isolated from wild-type or TLR2 knockout mice.
A survey of multiple batches of total lipid extracts from P. gingivalis, additional strains of P. gingivalis (W83 and 381), and several other members of the Bacteroidetes phylum, including two intestinal species, shows that Lipid 1256 is prevalent in these organisms relative to other previously characterized serine/glycine lipids. Preliminary results using extracted human teeth indicate that Lipid 1256 is specifically deposited on periodontally diseased teeth but not on impacted third molars that are surgically removed from the oral cavity (data not shown). Lipid 1256 levels are also elevated in lipid extracts from gingival-tissue surgical samples taken from periodontitis sites compared with healthy sites (data not shown). The local uptake of Lipid 1256 into tissues from subgingival plaque where Bacteroidetes organisms are prevalent or blood contamination with subsequent deposition in remote tissues could account in part for the reported role of P. gingivalis in promoting systemic disease (30). This issue is also potentially important for those Bacteroidetes organisms residing in the GI tract (31, 32) because they typically represent up to half of the intestinal microorganisms in humans (33). Because of the ability of Lipid 1256 to engage TLR2, chronic inflammatory bowel disease could be promoted or intensified through long-term penetration of this lipid into the bowel wall. Because Lipid 1256 is lipophilic, the contamination of blood and subsequent deposition of Lipid 1256 in artery walls and the brain, with long-term activation of TLR2, would be an important reason for further investigation of the recovery of this lipid in human tissue samples. In summary, Lipid 1256 could eventually be shown to participate in a variety of chronic inflammatory conditions that are associated with the pathological engagement of TLR2. Further research is required to understand the biological processes involved with the synthesis of Lipid 1256 and the relevance of this novel bacterial lipid class to TLR2-associated chronic inflammatory diseases in humans, such as atherosclerosis and autoimmune disease.
In conclusion, the current investigation has identified a new lipid class of P. gingivalis, Lipid 1256, that appears to be a relatively strong ligand for TLR2. This novel lipid class consists of a phosphoglycerol serine-glycine lipodipeptide substituted with four fatty acids. The fatty acids in Lipid 1256 are predominantly iso-branched with lesser amounts of anteiso-branched fatty acids, and the branched fatty acids may influence the biological activity of this lipid class, as was reported for the phosphatidylethanolamine lipid class of P. gingivalis (29). It is not clear whether similar complex lipids produced by P. gingivalis can be modified with acylated glycerol phosphate. For example, the serine-glycine lipodipeptide of P. gingivalis, Lipid 430, could also be substituted on its terminal serine with diacylated phosphoglycerol in a manner similar to Lipid 654. Future work will investigate this possibility.

Data availability
HPLC, MS, and NMR data are included in the article. NMR chemical shifts were deposited in the Biological Magnetic Resonance Data Bank under BMRB ID 50303. The HEK cell results, monocyte secretory responses, and bacterial lipid recovery data are available on request from the corresponding author.