Porphyromonas gingivalis Lipids Inhibit Osteoblastic Differentiation and Function

Abstract

Porphyromonas gingivalis produces unusual sphingolipids that are known to promote inflammatory reactions in gingival fibroblasts and Toll-like receptor 2 (TLR2)-dependent secretion of interleukin-6 from dendritic cells. The aim of the present study was to examine whether P. gingivalis lipids inhibit osteoblastic function. Total lipids from P. gingivalis and two fractions, phosphoglycerol dihydroceramides and phosphoethanolamine dihydroceramides, were prepared free of lipid A. Primary calvarial osteoblast cultures derived from 5- to 7-day-old CD-1 mice were used to examine the effects of P. gingivalis lipids on mineralized nodule formation, cell viability, apoptosis, cell proliferation, and gene expression. P. gingivalis lipids inhibited osteoblast differentiation and fluorescence expression of pOBCol2.3GFP in a concentration-dependent manner. However, P. gingivalis lipids did not significantly alter osteoblast proliferation, viability, or apoptosis. When administered during specific intervals of osteoblast growth, P. gingivalis total lipids demonstrated inhibitory effects on osteoblast differentiation only after the proliferation stage of culture. Reverse transcription-PCR confirmed the downregulation of osteoblast marker genes, including Runx2, ALP, OC, BSP, OPG, and DMP-1, with concurrent upregulation of RANKL, tumor necrosis factor alpha, and MMP-3 genes. P. gingivalis total lipids and lipid fractions inhibited calvarial osteoblast gene expression and function in vivo, as determined by the loss of expression of another osteoblast differentiation reporter, pOBCol3.6GFPcyan, and reduced uptake of Alizarin complexone stain. Finally, lipid inhibition of mineral nodule formation in vitro was dependent on TLR2 expression. Our results indicate that inhibition of osteoblast function and gene expression by P. gingivalis lipids represents a novel mechanism for altering alveolar bone homeostasis at periodontal disease sites.

Bone loss under the influence of bacterial virulence factors is thought to occur through engagement of receptors for pathogen-associated molecular pattern (PAMP) molecules, resulting in the stimulation of osteoclasts and/or the inhibition of osteoblasts. A frequently cited example is the engagement of Toll-like receptor 4 (TLR4) by bacterial lipopolysaccharide (LPS) that is reported to mediate bone loss in destructive periodontal diseases through activation of osteoclasts and inhibition of osteoblasts (9, 21, 27, 40, 44, 45). A subgingival organism strongly associated with destructive periodontal disease, Porphyromonas gingivalis, is a Gram-negative anaerobe that produces LPS, but this LPS is unique in that it has been shown to engage both TLR4 and TLR2 (4, 10, 20, 24, 37), although these reports are sometimes conflicting. Though inflammatory bone loss in experimental periodontal diseases can be produced by engagement of TLR4 by LPS (28), recent evidence indicates that periodontal bone loss in experimental animals mediated by live P. gingivalis requires engagement of TLR2 (16). Furthermore, previous reports have shown that LPS from P. gingivalis is present only to a negligible extent in diseased periodontal tissues in humans (30, 33), and the lipid A of P. gingivalis cannot be detected on periodontally diseased teeth or in diseased periodontal tissues (35). In contrast, unusual complex sphingolipids of P. gingivalis, called phosphoethanolamine dihydroceramide (PE DHC) or phosphoglycerol dihydroceramide (PG DHC) lipids, are prominent on periodontally diseased teeth and are recovered in diseased periodontal tissues (35). We recently reported that the total lipids of P. gingivalis, and specifically the PE DHC lipids, stimulate interleukin-6 (IL-6) secretion from dendritic cells in vitro, and this process is mediated through TLR2 (31). The present investigation sought to determine whether the lipids of P. gingivalis could potentially interfere with bone formation in periodontal disease sites through the inhibition of osteoblast function or its phenotype and whether this process requires engagement of TLR2.

MATERIALS AND METHODS

Preparation and purification of P. gingivalis lipids.

P. gingivalis (ATCC 33277, type strain) was inoculated into basal (peptone, Trypticase, and yeast extract) medium supplemented with hemin and menadione (Sigma, St. Louis, MO) and brain heart infusion (34). Culture purity was verified by Gram staining, the lack of growth in aerobic culture, and the formation of black colonies when inoculated on blood agar plates and grown under anaerobic conditions. The suspension cultures were incubated in an anaerobic chamber flushed with N₂ (80%), CO₂ (10%), and H₂ (10%) at 37°C for 5 days, and the bacteria were harvested by centrifugation (3,000 × g for 20 min). After lyophilization, ∼4 g of P. gingivalis pellet was extracted for 5 days by using a modification of the phospholipid extraction procedure of Bligh and Dyer (7) and Garbus et al. (12). Specifically, 4 ml of H₂O plus 16 ml of MeOH-CHCl₃ (2:1 [vol/vol]) were added to the bacterial sample, and the mixture was vortexed. After 12 h, 3 ml of 2 N KCl plus 0.5 M K₂HPO₄ and 3 ml of CHCl₃ were added, and the sample was vortexed. The lower organic phase was carefully removed, and CHCl₃ (3 ml) was added to each sample and vortexed. The CHCl₃ phase was removed and combined with the previous organic solvent extract. The lipid extract from P. gingivalis was dried under nitrogen and stored frozen. This lipid preparation was used as the total lipid extract for the experiments described below.

Fractionation of bacterial lipids by high-performance liquid chromatography (HPLC) was accomplished by using a semipreparative HPLC column (1-cm-by-25-cm silica gel, 5 mm; Supelco, Inc., Bellefonte, PA) and eluted isocratically with hexane-isopropanol-water (6:8:0.75 [vol/vol/vol] solvent A) (13). P. gingivalis lipid samples were dissolved in solvent A and centrifuged to remove insoluble material. For each chromatographic separation, 20 mg of lipid was applied, and fractions were pooled from 20 column fractionations. Lipids were eluted at 2.0 ml/min and monitored at 205 nm with fractions collected every minute. Fractions were dried under nitrogen and resuspended in 2 ml of CHCl₃. Lipid recovery in each HPLC fraction were determined by drying 5 μl from each fraction onto a microbalance tray and weighing the tray using a Cahn Electrobalance. Fractions shown to be enriched for PE DHC or PG DHC lipids by electrospray-mass spectrometry (MS) (see below) were pooled, dried, and refractionated. The refractionated lipid fractions were then verified by electrospray-MS to be highly enriched as shown in Fig. 1.

FIG. 1.

Electrospray-MS characterization of total lipids of P. gingivalis and the highly enriched phosphoethanolamine dihydroceramide (PE DHC) or phosphoglycerol dihydroceramide (PG DHC) lipid fractions. Lipids were prepared and characterized by electrospray-MS as described in Materials and Methods. The upper frame shows the negative ions recovered in the total lipid extract of P. gingivalis lipids, the middle frame shows the ions recovered in the PE DHC lipid fraction, and the bottom frame shows the ions recovered in the PG DHC lipid fraction. A description of the MS interpretation of these ions and the structural deduction can be found in Nichols et al. (34). The PG DHC lipids include three component lipids that produce negative ions of 960, 946, and 932 m/z and, by analogy, the PE DHC lipids produce negative ions of 705, 691, and 677 m/z (34). Note that the dominant-negative ions previously reported for lipid A species of P. gingivalis (1,450, 1,690, and 1,770 m/z) (10) are not recovered in these lipid isolates.

The lipid fractions were surveyed by using electrospray-MS analysis on a Micromass Quattro II mass spectrometer system with a low-flow HPLC introduction system as previously described (34). Dihydroceramide lipid fractions were dissolved in hexane-isopropanol (6:8 [vol/vol], elution solvent) and were applied at a flow rate of 80 μl/min. For electrospray negative-ion analyses, the desolvation and inlet block temperatures were 80 and 100°C, respectively, and the transcapillary potential was 3,000 V. The cone voltage was usually held at 30 V, and the mass acquisition range was 100 to 2,000 amu for initial electrospray-MS analyses. The lipid preparations used for the present study were shown to be devoid of lipid A of P. gingivalis by negative-ion electrospray-MS (Fig. 1). Negative ions characteristic for lipid A species of P. gingivalis were not observed in any of the lipid preparations shown in Fig. 1 (10).

Primary calvarial osteoblast cultures.

Primary cell cultures were established from calvarial cells isolated from 5- to 7-day-old neonatal CD-1 mice or TLR2^−/− mice (a generous gift from S. Akira, Osaka, Japan). Mice were maintained and bred in accordance with University of Connecticut Center for Laboratory Animal Care regulations. Briefly, the calvaria were isolated from skulls without sutures, and adherent mesenchymal tissues and osteoblasts were released by subjecting these calvaria to four sequential 15-min enzyme digestions at 37°C in a solution containing 0.05% trypsin-EDTA and 0.1% collagenase P (Roche Diagnostics, Indianapolis, IN). Cells released from the second to fourth digestions were collected, centrifuged, resuspended, and plated at 1.5 × 10⁴ cells/cm² (i.e., 1.5 × 10⁵ cells/well) in 35-mm six-well culture plates in Dulbecco modified Eagle medium (DMEM; Invitrogen, Carlsbad, CA) containing 10% fetal calf serum, penicillin (100 U/ml), streptomycin (100 μg/ml), and nonessential amino acids (100 μM). The day of plating was designated day 0. Plated cells became confluent around day 5 to 6 at which point the culture medium was changed to differentiation medium, which was α-MEM (Invitrogen) containing 10% fetal calf serum, penicillin (100 U/ml), streptomycin (100 μg/ml), ascorbic acid (50 μg/ml), and β-glycerophosphate (4 mM). Medium was changed every other day for the entire duration (21 days) of culture. For the evaluation of osteoblast differentiation using fluorescence microscopy, calvarial osteoblast cultures were derived from mice transgenic for pOBCol2.3GFP (22) in which the green fluorescence protein (GFP) reporter was driven by a rat 2.3-kb type I collagen promoter. For the evaluation of osteoblast engagement of TLR2 by P. gingivalis lipids, calvarial osteoblast cultures were derived from TLR2 knockout mice that were backcrossed 9 times and verified by genotyping to be homozygous for TLR2^−/−. TLR2^−/− osteoblast cultures were otherwise handled using the same culture protocol as described for CD-1 osteoblasts.

Assessment of mineralized nodule formation.

Mineralization nodules were assessed by using the modified von Kossa's silver nitrate staining method. Briefly, cultures were fixed in cold methanol for 15 to 20 min. After a rinsing step, the fixed plates were incubated with a 5% silver nitrate solution under UV light using two cycles of auto-cross-linking (1,200 μJ × 100) in a UV Stratalinker (Stratagene, La Jolla, CA). Mineralized nodules were seen as dark brown to black spots. Culture plates stained with von Kossa were scanned, and the areas of mineralized nodules were quantitated by using a computation program developed from Openlab (Improvision, Lexington, MA). The area of mineralized nodule formation was quantified in pixels.

Examination of cell viability and apoptosis.

Ethidium homodimer-1 (EthD-1; Molecular Probes, Eugene, OR) was used to determine cell viability. EthD-1, with a high affinity for DNA and low membrane permeability, permits the use of a very low fluorescent dye concentration to demonstrate cell death. Osteoblast cultures were incubated with EthD-1 at a final concentration of 2 μM for 30 min and were examined for cell viability by fluorescence microscopy using a TRITC (tetramethyl rhodamine isothiocyanate) filter. The fluorescence image of EthD-1 staining (red color) was photographed, and cell death was quantified by calculating the total number of cells with red stain divided by the total number of cells in the field. To examine apoptosis, allophycocyanin (APC)-conjugated annexin V was used with flow cytometry to detect apoptosis. Annexin V has a high affinity for phosphatidylserine that is translocated to the outer surface of the cell membrane during apoptosis. Harvested cultured cells were resuspended in annexin binding buffer which consisted of 10 mM HEPES, 140 mM NaCl and 2.5 mM CaCl₂ at pH 7.4. APC-conjugated annexin V (Molecular Probes) was added to the cell suspension at 1:20 dilution to label the apoptotic cells for 15 min. Cells were analyzed by flow cytometry using channel FL4 to detect the apoptotic cells labeled with APC-conjugated annexin V.

RT-PCR analysis of gene expression.

Total RNA was extracted from cultures by using TRIzol reagent (Invitrogen) and phenol-chloroform according to the manufacturer's instructions. RNA was dissolved in Tris-EDTA (pH 7.4), and the concentration of RNA was determined by absorbance at 260 nm. RNA was treated with DNase I (Invitrogen) to remove genomic DNA contaminants. Reverse transcription (RT) was carried out in a 20-μl volume containing ∼3 μg of RNA, 1 μl of random hexamers (50 ng/μl), 1 μl of annealing buffer, 10 μl of 2× first-strand reaction mix, and 2 μl of Superscript III/RNase OUT enzyme mix (Invitrogen) at 25°C for 10 min, followed by 50°C for 50 min. TaqMan real-time PCR was performed from 1 μl of cDNA using TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA) with 100 nM primers and 50 nM probe. The TaqMan real-time PCR was performed on a TaqMan ABI 7500 system (Applied Biosystems). The unlabeled specific primers and the TaqMan MGB probes (6-FAM dye-labeled) for gene detection were as follows: mouse Runx2 (assay ID, Mm00501580_m1), type I collagen (Col1a1; Mm00483888_ml), bone alkaline phosphatase (ALP; Mm01187117_m1), bone sialoprotein (BSP; Mm00492555_m1), osteocalcin (OC; Mm00649782_9H), dentin matrix protein-1 (DMP-1; Mm01208365_m1), matrix metalloproteinase-3 (MMP-3; Mm01168406_g1), osteoprotegerin (OPG; Mm00435454-m1), tumor necrosis factor alpha (TNF-α; Mm00443258_m1), and receptor activator of NF-κB ligand (RANKL; Mm01313944_g1). GAPDH was used as an internal control (Mm99999915_g1). Cycling conditions were as follows: after an initial hold of 2 min at 50°C and 10 min at 95°C, the samples were cycled 40 times at 95°C for 15 s and 60°C for 1 min. Each sample was assayed in triplicate. The comparative threshold cycle value (C_T) method was applied to determine the comparative expression levels between samples relative to control gene expression. To examine regulation by lipids, the amplification C_T from the lipid treated samples were subtracted from the untreated sample cycle values (ΔC_T = C_{T untreated} − C_{T treated}). The ratio was obtained by calculating the values of each gene of interest against the house-keeping gene GAPDH. The fold change of the test gene was determined as 2^{(ΔC_T gene − ΔC_T GAPDH)}.

In vivo inhibition of osteoblasts.

Mice were bred to express the osteoblastic FP reporter, pOBCol3.6GFPcyan, on a CD-1 background as previously described (6). Lipid preparations, including total lipids, PE DHC lipids and PG DHC lipids, were sonicated (2 min, 3 W) in phosphate-buffered saline (PBS) to achieve a concentration of 0.1 μg/ml. Mice were lightly sedated with isoflurane, and lipid preparations were administered subcutaneously to the calvaria midway between the ear and eye. The right calvarium of each mouse received a subcutaneous injection of 5 μg of lipid. Control mice calvaria received only PBS. Mice were administered Alizarin complexone for labeling of newly deposited mineralized tissue during day 6 and were sacrificed on day 7. Calvaria were cryosectioned and photographed for osteoblast reporter (blue) and Alizarin complexone (red) by fluorescence microscopy. Next, sections were stained with hematoxylin and eosin for histology.

Statistical analysis.

All cell culture experiments were analyzed by using either the Student t test or one-factor analysis of variance, followed by post-hoc testing for differences between treatment categories (Scheffe contrasts among pairs of means). Values were presented as mean ± the standard error unless noted otherwise. Differences were considered statistically significant at P < 0.05.

RESULTS

Effects of P. gingivalis lipids on mineralized nodule formation in cultures.

Calvarial osteoblast cultures derived from 5- to 7-day-old mice were used to examine the effect of P. gingivalis total lipids on mineralized nodule formation. The concentration of P. gingivalis total lipids in culture medium was varied from 2,500 to 156 ng/ml. Lipids were suspended in culture medium at the highest concentration by sonication and subsequently diluted in fresh medium to achieve 2-fold serial dilutions. Mineralized nodule formation revealed by von Kossa staining in day 21 cultures showed that P. gingivalis total lipids inhibited mineral deposition in a concentration-dependent manner (Fig. 2A). Although the lowest concentration of 156 ng/ml had no significant effect, the highest concentration of 2,500 ng/ml completely inhibited the mineralized nodule formation.

FIG. 2.

Effects of P. gingivalis total lipids on the mineralized nodule formation and osteoblastic differentiation at day 21 calvarial osteoblast cultures. Cells were cultured with increasing concentrations of P. gingivalis total lipids for the 21-day culture period. Control cultures were not exposed to bacterial lipids. (A) Mineralized nodule formation revealed by von Kossa staining showed that the inhibitory effect of P. gingivalis total lipids was concentration dependent. (B) The inhibition of osteoblastic differentiation revealed by fluorescence of pOBCol2.3GFP showed that the inhibitory effect of P. gingivalis total lipids was also concentration dependent. Scale bar, 10 mm. (C) Quantitation of mineralized nodule formation or fluorescence expression of pOBCol2.3GFP in control osteoblast cultures or cultures exposed to increasing levels of P. gingivalis total lipids. Both von Kossa staining and GFP expression were quantified by summing pixels indicating either von Kossa staining or GFP fluorescence. The results in Fig. 2C represent the mean ± the standard deviation for n = 2 trials. One-factor analysis of variance showed significant differences between the treatment categories: von Kossa staining (P, 0.00181) and GFP fluorescence (P, 0.00007).

To further examine the inhibitory effect of P. gingivalis total lipids on osteoblastic differentiation in cultures, transgenic mice expressing the osteoblast-specific pOBCol2.3GFP marker, which is driven by a 2.3-kb fragment of the rat Col1a1 promoter, were used to establish calvarial osteoblast cultures. Previous studies have demonstrated that the expression of pOBCol2.3GFP is observed in differentiated osteoblasts and osteocytes at the time when bone sialoprotein and osteocalcin are strongly expressed in cultures (22). Consistent with mineral nodule deposition in Fig. 2A, the highest concentration of lipids (2500 ng/ml) had a strong inhibitory effect on the expression of pOBCol2.3GFP in day 21 cultures (Fig. 2B). When quantified using replicate cultures, the dose responses were observed to be highly significant for both von Kossa staining or GFP fluorescence (Fig. 2C). These results show that P. gingivalis total lipids inhibit mineralized nodule formation and osteoblast differentiation in a concentration-dependent manner. For the remaining experiments of the present study, calvarial osteoblasts cultures were treated with 1,250 ng/ml of P. gingivalis total lipids or lipid fractions.

Effects of P. gingivalis lipids on cell viability, apoptosis, and proliferation.

To test whether P. gingivalis total lipids affect cell viability, EthD-1, a red-fluorescent indicator of cell death, was added to the culture medium, and EthD-1-positive cells were quantified by fluorescence microscopy. Lipid-treated and control cultures (126.5 ± 15.6 versus 107.6 ± 12.0, respectively; P = 0.35) (Fig. 3A) showed no significant difference in the percentage of cells staining for EthD-1. Next, annexin V staining was used to examine for apoptosis in cultures. Apoptotic cells labeled with fluorochrome-conjugated annexin V were quantitated by fluorescence-activated cytometric analysis. Cells harvested from cultures treated with or without P. gingivalis total lipids were similar in apoptotic rates (5.9% ± 0.5% versus 6.3% ± 0.5%, respectively; P = 0.59) (Fig. 3B).

FIG. 3.

Effects of P. gingivalis total lipids on cell viability, apoptosis, and proliferation. The concentration of P. gingivalis total lipids was 1,250 ng/ml in the indicated cultures. (A) Cell viability indicated by EthD-1 staining at day 7 showed no significant difference in cell death between lipid-treated and control osteoblast cultures (126.5 ± 15.6 versus 107.6 ± 12.0; P = 0.35). (B) Apoptosis observed at day 7 with annexin V staining, followed by flow cytometry, showed no significant difference between lipid-treated and control osteoblast cultures (5.9% ± 0.5% versus 6.3% ± 0.5%; P = 0.59). (C) Total cell counts per well at day 7 showed no significant difference in total cell recovery between lipid-treated and control osteoblast cultures ([306.6 ± 15.8] × 10³ versus [263.0 ± 25.3] × 10³; P = 0.18). (D) DNA quantitation showed no significant differences in total DNA recovered from control and lipid treated osteoblasts for day 7 cultures (4.3 ± 0.2 μg/ml versus 4.5 ± 0.5 μg/ml; P = 0.79), as well as day 21 cultures (8.5 ± 0.3 μg/ml versus 8.3 ± 0.4 μg/ml; P = 0.71).

The effect of P. gingivalis total lipids on cell proliferation was evaluated by quantifying cell number and DNA. At day 7 of culture, no difference in cell counts was observed between lipid-treated and control cultures ([306.6 ± 15.8] × 10³ versus [263.0 ± 25.3] × 10³ per well, respectively; P = 0.18) (Fig. 3C). Moreover, at both day 7 and day 21 of culture, there was no significant difference in total cell DNA between the lipid-treated and control cultures (Fig. 3D). In summary, P. gingivalis total lipids did not affect cell proliferation, necrosis, or apoptosis in calvarial osteoblast cultures despite the strong inhibitory effects on osteoblastic differentiation and mineralized nodule formation. Given that P. gingivalis total lipids have no significant effect on cell proliferation or apoptosis, P. gingivalis lipid inhibition of osteoblast differentiation and mineral nodule formation in culture cannot be explained by selective cell death or apoptosis or cell proliferation.

Effects of P. gingivalis lipids on osteoblast differentiation in vitro.

The 3-week period of calvarial osteoblasts cultures used in the present study can be generally divided into three major in vitro developmental stages: proliferation (first week), differentiation (second week), and mineralization (third week) (Fig. 4A). To examine the effects on different stages of growth in cultures, P. gingivalis total lipids were added during the first week only, the second week only, the third week only, both the first and the second weeks, both the second and the third weeks, or continuously for all 3 weeks (Fig. 4B). The results of von Kossa staining at day 21 showed that the exposure to total lipids during only the first week did not affect mineralized nodule formation by 3 weeks of culture compared to controls. However, the administration of lipids during either the second week only or the third week only significantly inhibited the mineralized nodule formation (Fig. 4B). Interestingly, cultures exposed to lipids for the first through the second week of culture, but not the second through the third week of culture, showed intermediate mineral nodule formation similar to cultures exposed to lipids either during the second week or third week of culture. Cultures exposed to lipids during the second and third weeks, as well as cultures continuously exposed to lipids for 3 weeks, showed the least mineralized nodule formation (Fig. 4B).

FIG. 4.

Inhibition of the mineralized nodule formation by P. gingivalis total lipids depending on the week of lipid treatment in culture. (A) Cells were exposed to P. gingivalis total lipids (1,250 ng/ml) only during the specified culture intervals, including the first week only, the second week only, the third week only, the first and second week, the second and third week, or all 3 weeks. Control cultures were not exposed to P. gingivalis lipids. (B) Quantitation of mineralized nodule formation stained by von Kossa of day 21 cultures. The magnitude of inhibitory action of P. gingivalis total lipids was dependent on the stage of growth in cultures. Control cultures and cultures treated for the first week with lipids (&) were not significantly different from each other but both were significantly different from all other treatment groups, as determined by Scheffe contrasts among pairs of means. Cultures treated with lipids for the second week, the third week, or the combined first and second weeks (*) were not significantly different from each other but were significantly different from the remaining cell culture categories. Cultures treated for the second and third weeks or for all weeks (#) were not significantly different from each other but were significantly different from the remaining cell culture categories.

Effects of P. gingivalis lipids on gene expression.

To examine the gene expression effects of P. gingivalis total lipids, total RNA was isolated from day 21 osteoblast cultures after continuous exposure to lipids (1250 ng/ml) and analyzed by real-time RT-PCR. Genes selected for RT-PCR assessment were those involved with osteoblast differentiation, secretion, and/or mineralization, including Runx2, ALP, BSP, OC, Col1a1, and DMP-1 genes, and those involved with matrix remodeling (MMP-3) or indirect activation of osteoclastogenesis, including OPG, TNF-α, and RANKL genes.

RT-PCR results showed that the expression of ALP, OC, Col1a1, and Runx2 were generally downregulated by P. gingivalis lipids compared to the control cultures (Fig. 5). In particular, BSP and DMP-1 showed the strongest downregulation (Fig. 5). Decreased expression of osteoblast genes directly correlated with the observed effects of P. gingivalis lipids on osteoblast differentiation as shown above by decreased pOBCol2.3GFP expression and reduced mineralized nodule formation revealed by von Kossa staining.

FIG. 5.

Effects of P. gingivalis total lipids on gene expression determined by real-time RT-PCR. Cells were cultured with P. gingivalis total lipids (1,250 ng/ml) or control medium for the entire 21-day culture period. Real-time RT-PCR was performed on total RNA isolated from day 21 cultures. Changes in gene expression, either as increased or decreased gene expression, in lipid-treated cultures are expressed as the fold change versus control cultures. Significant up or down expression of each gene was evaluated against parallel control cultures by using the Student t test, and P values are depicted opposite each histogram bar. Only the expression of the Col1a1 and RANKL genes was not significantly affected by lipid treatment of osteoblast cultures.

Because bone formation is coupled to bone resorption through cell surface expression of RANKL by osteoblasts or the release of soluble mediators from osteoblasts, we also examined the expression of osteoblast genes capable of activating osteoclasts. Our results showed that the exposure of calvarium-derived osteoblasts to P. gingivalis total lipids increased the expression of genes for factors known to activate osteoclasts, including TNF-α and RANKL (Fig. 5). Expression of MMP-3 was also upregulated consistent with increased bone matrix remodeling (Fig. 5). Together, these in vitro results indicate that P. gingivalis total lipids can promote bone loss by directly inhibiting osteoblast differentiation and suggest indirect activation of osteoclast-mediated bone resorption through elevated expression of genes capable of promoting osteoclast activation.

Effects of P. gingivalis total lipids and lipid fractions on calvarial osteoblasts in vivo.

In Fig. 2, a dose-dependent reduction in mineral nodule deposition was observed to coincide with diminished expression of the osteoblastic GFP reporter, pOBCol2.3GFP. The pOBCol2.3GFP transgene is reported to be a late osteoblast differentiation marker. We next sought to examine bacterial lipid inhibition of osteoblasts in vivo using the pOBCol3.6cyan (blue) transgene model that reveals osteoblasts in both the early and late differentiation states. The marker pOBCol3.6GFPcyan is driven by a 3.6-kb fragment of the rat Col1a1 promoter (6). Histological analysis has shown that this transgene is active in periosteal fibroblasts and bone lining cells thus marking both preosteoblasts and differentiated osteoblasts. In primary osteoblast cultures derived from neonatal calvarial or bone marrow stromal fibroblasts, pOBCol3.6GFPcyan expression parallels the expression of alkaline phosphatase and Col1a1 mRNA during the first week in culture. However, the pOBCol3.6GFPcyan signal intensifies with the initiation of mineral nodule deposition, indicating osteoblast maturation.

For this evaluation, P. gingivalis lipid preparations were sonicated in PBS and administered subcutaneously to the calvaria of mice. As shown in Fig. 6, administration of bacterial lipid preparations to mouse calvaria in vivo substantially reduced blue fluorescence after 1 week and only along the external surface of the calvaria that were exposed to lipids. In addition, an Alizarin complexone fluorescent stain for bone mineral deposition was diminished in the area of inhibited osteoblast function. These results indicate that the total lipids of P. gingivalis, as well as the PE DHC and PG DHC fractions, significantly inhibit functional osteoblast activity in vivo as determined by essentially complete loss of differentiated osteoblasts on the surfaces of the calvaria, as measured by pOBCol3.6GFPcyan fluorescence, and substantial inhibition of bone mineral deposition, as measured by decreased Alizarin complexone staining. Whether this effect is due to dedifferentiation of osteoblasts in vivo or other effects remains to be established.

FIG. 6.

Effects of P. gingivalis lipid preparations on osteoblast function and gene expression in vivo. Mice that express the osteoblastic GFP reporter, pOBCol3.6GFPcyan, were lightly sedated, and the indicated lipid preparations (5 μg in 50 μl of PBS) were administered as a subcutaneous injection to the surfaces of the calvaria (two mice were treated with each lipid preparation, including the total lipid extract, PG DHC lipids and PE DHC lipids of P. gingivalis). Control mouse calvaria received only PBS. Mice were administered Alizarin complexone on day 6 and were sacrificed on day 7. Calvaria were cryosectioned and evaluated by fluorescence microscopy for osteoblast reporter (blue) and alizarin complexone (red) fluorescence (upper row). The right side of each calvarium section represents the cerebral cavity, and the left side represents the surface treated with the indicated lipid preparation. The inserts within each frame of the upper row depict magnified images of lipid-treated (left) and untreated (right) surfaces of each calvarium section. The exact same sections depicted in the upper row were then stained with hematoxylin and eosin (H&E), and photomicrographs were obtained (lower row). Multiple sections of each calvarium specimen were evaluated for fluorescence changes, and the images depicted above are representative of those observed in the replicate sections.

P. gingivalis lipids mediate effects on calvarial osteoblasts through TLR2.

We next examined the role of TLR2 in lipid-induced inhibition of osteoblast differentiation. The effects of P. gingivalis total lipids and lipid fractions were examined in cultures derived from either TLR2 knockout (TLR2^−/−) mice or wild-type (WT) CD-1 mice. Calvarial osteoblast cultures derived from both WT and TLR2^−/− mice reached similar levels of mineralized nodule formation by day 21 (Fig. 7, controls). Mineralized nodule formation was substantially inhibited by P. gingivalis total lipids in the WT cultures. In contrast, P. gingivalis total lipids did not inhibit mineral nodule formation in TLR2^−/− osteoblasts (Fig. 7). For comparison, we also examined the effect of a TLR4 ligand, Salmonella enterica serovar Typhimurium LPS (1 μg/ml; Difco Laboratories, Detroit, MI), on mineral nodule formation in osteoblasts derived from either WT or TLR2^−/− mice. LPS effectively suppressed mineralized nodule formation in osteoblasts derived from both WT and TLR2^−/− mice (Fig. 7). Finally, we evaluated the two major fractions of P. gingivalis lipids, PG DHC and PE DHC lipids, for inhibition of mineralized nodule formation. PG DHC lipids substantially reduced mineralized nodule formation in WT osteoblasts, whereas PE DHC lipids did not have an inhibitory effect (Fig. 7, PG DHC and PE DHC). Furthermore, PG DHC treatment did not inhibit mineralized nodule formation in osteoblasts derived from TLR2^−/− mice compared to control osteoblasts, indicating that PG DHC lipids act through TLR2. Our results indicate that the inhibitory effect of P. gingivalis lipids on osteoblast differentiation is mediated by TLR2 and the highly enriched PG DHC lipid fraction, but not PE DHC lipid fraction, is a primary lipid fraction responsible for this inhibitory effect.

FIG. 7.

Engagement of TLR2 by P. gingivalis total lipids and major fractions of P. gingivalis lipids. Calvarial osteoblast cultures were established from the WT CD-1 or TLR2-null (TLR2^−/−) mice and treated with P. gingivalis total lipids, LPS, or two major P. gingivalis lipid fractions (PG DHC or PE DHC lipids) for the entire 21-day culture period. The von Kossa staining was performed at day 21 to reveal mineralized nodule formation in cultures. Control cultures did not receive P. gingivalis lipids, LPS, or lipid fractions. The inhibitory action of P. gingivalis total lipids was TLR2 dependent, and the PG DHC fraction, but not the PE DHC fraction, demonstrated TLR2-dependent inhibition of osteoblast mineralized nodule formation.

DISCUSSION

The present study demonstrates that P. gingivalis total lipids and two previously reported, biologically active lipid fractions, specifically the PG DHC and PE DHC lipid fractions (34), affect osteoblast differentiation and/or gene expression. We have previously demonstrated that lipid extracts from teeth with gross amounts of subgingival calculus contain essentially all lipid ions present in the total lipid extract of P. gingivalis (35). Because diseased gingival tissues are exposed to total lipids of P. gingivalis through contact with subgingival calculus and previous reports suggest that P. gingivalis can invade gingival tissues (25, 41, 42), the total lipid extract of P. gingivalis should be the principal lipid preparation for evaluating effects on bone cells. Regarding specific lipid classes of P. gingivalis, PG DHC lipids were considerably more abundant on periodontally diseased teeth than the PE DHC lipids, and neither of these lipid classes were detected in lipids extracted from an impacted third molar (35). Of note, the common lipid A species (1, 10) of P. gingivalis were not detected on the diseased root surfaces (35). Lipid extracts of diseased gingival tissue samples also contained PE DHC and PG DHC lipids (35), and these lipids are known to produce strong proinflammatory responses in fibroblasts (34) or dendritic cells in vitro (31). Recently, lipid extracts of six diseased gingival tissue samples were individually analyzed for P. gingivalis lipids using multiple reaction monitoring MS/MS (unpublished data), and this analysis confirmed the previously reported recovery of PG DHC and PE DHC lipids in diseased gingival tissues. At this point, the PE DHC and PG DHC lipids are the only bacterial lipid classes that we have observed in lipid extracts of diseased gingival tissues. For this reason and the reasons cited above, our investigation was limited to the evaluation of the total lipid extract of P. gingivalis and the PE DHC and PG DHC lipid fractions.

Although others have suggested that P. gingivalis can mediate its effects on alveolar bone destruction through engagement of other TLRs (5, 28), several virulence factors of P. gingivalis have been implicated as TLR2 ligands, including lipoprotein (2), fimbriae (3, 17, 18, 38), and LPS (4, 10). One or more of these factors derived from P. gingivalis could account for TLR2-dependent stimulation of alveolar bone loss (16, 43) or other pathogenic processes, including experimental atherosclerosis (15, 26) observed after P. gingivalis infection. However, aside from lipoprotein (39), most evidence shows that LPS (24) or fimbriae (11, 19, 38) from P. gingivalis have limited capacity to engage TLR2 or only partially engage TLR2. We previously reported that PE DHC lipids of P. gingivalis lipids stimulate IL-6 release from dendritic cells in a TLR2-dependent manner (31). Live P. gingivalis is reported to stimulate release of TNF-α, but not RANKL, from peritoneal macrophages, and this macrophage activation requires engagement of TLR2 (43). The TLR2-dependent release of TNF-α mediated by live P. gingivalis was also shown to be critical for macrophage stimulation of osteoclastogenesis (43). The results of our study indicate that complex lipids of P. gingivalis could mediate TLR2-dependent alveolar bone loss at periodontal disease sites by directly inhibiting osteoblast activity. Furthermore, our study suggests that P. gingivalis lipids could indirectly stimulate osteoclast activity through elevated expression of osteoblast genes capable of stimulating osteoclasts or osteoclastogenesis, including increased TNF-α and RANKL expression and decreased expression of OPG. Whether P. gingivalis lipids directly promote osteoclast activation will be the subject of future investigation.

Osteoblast differentiation in vitro represents a series of proliferation and differentiation stages over the course of 21 days in culture and is dependent on specific medium supplements to induce osteoblast differentiation. The inhibition of osteoblast differentiation by P. gingivalis lipids becomes apparent in the second week of cell culture, and inhibition increases into the third week of osteoblast differentiation. Using similar culture conditions, a previous report demonstrated inhibition of osteoblasts in vitro by P. gingivalis LPS at levels slightly lower than the doses of P. gingivalis total lipids used in the present study. According to Kadono et al. (21), 100 ng of P. gingivalis LPS/ml reduced mineral nodule formation by 71%, whereas our study showed that 156 ng of P. gingivalis total lipids/ml reduced mineral nodule formation by 48% (Fig. 2). Of note, we have demonstrated by using electrospray-MS that our lipid fractions are not contaminated by LPS or lipid A constituents. However, LPS preparations of P. gingivalis used by other investigators are not established by MS methods to be free of contaminating lipids. The possibility of lipid contamination is now a concern for the evaluation of P. gingivalis LPS or lipid A in bone cell assays.

In contrast to the inhibition of osteoblast differentiation in vitro, inhibition of osteoblast function in vivo requires only the presence of the indicated P. gingivalis lipid preparations. Both the differentiated osteoblast phenotype, as determined by pOBCol3.6GFPcyan transgene fluorescence, and active bone mineral deposition, as measured by Alizarin complexone uptake on the surface of bone, are markedly inhibited by the total lipids of P. gingivalis, as well as the PE DHC and PG DHC lipid classes. Although the osteoblast function and gene expression are affected by the various lipid preparations, higher-magnification imaging of the hematoxylin-and eosin-stained sections shows that the lipid-treated bone surfaces are largely covered with lining cells (images not shown). In contrast to either the total lipid extract or PG DHC lipid preparations, PE DHC lipids only partially inhibited Alizarin complexone uptake. This suggests that active bone mineral deposition was not inhibited as strongly by the PE DHC lipids, although the pOBCol3.6GFPcyan transgene expression was essentially completely inhibited. Perhaps other osteoblast differentiation genes, including osteocalcin and alkaline phosphatase genes, are not affected by PE DHC lipids, and the sustained expression of these differentiation genes could partially account for the apparent continued bone mineral deposition in the presence of PE DHC lipids. Also possible is that PE DHC lipids affect osteoblasts through different signaling pathways than the PG DHC lipids. These issues will be the subject of future investigation.

The effects of the P. gingivalis lipid preparations in vivo contrast with reports of LPS effects on calvarial bone after subcutaneous administration of LPS in vivo. We observed impressive inhibition of pOBCol3.6GFPcyan expression after only 1 week of exposure to 5-μg P. gingivalis lipid preparations. Subcutaneous administration of LPS, either as a single injection of 500 μg (8) or replicate injections of 250 μg (29), was required to significantly increase osteoclast activity in mouse calvaria in vivo. Although lipid effects on osteoblast function in vivo are not directly comparable to LPS effects in vivo, our results clearly show that the effect of P. gingivalis lipids on calvarial osteoblasts occurs at relatively low doses compared to the LPS levels required to promote osteoclast-activated bone resorption in vivo. Therefore, inhibition of osteoblast function and gene expression promoted by P. gingivalis lipids will likely result in net bone loss even without stimulation of bone resorption.

TLR2-dependent inhibition of osteoblast activity in vitro by PG DHC but not by PE DHC lipids is likely related to the structural differences between these lipid classes (see Fig. 1). The most significant structural difference is the ethanolamine versus the glycerol head group (34). However, the number of aliphatic chains may be equally if not more important than the head group in promoting the TLR2-dependent inhibition of osteoblast activity. The PG DHC lipids include three aliphatic chains, two of which are fatty acids with terminal branched aliphatic chains (isobranched), and the third aliphatic chain is part of the dihydroceramide long chain base structure (Fig. 1) (34). As with all dihydroceramide lipids produced by P. gingivalis, the PG DHC lipid class includes three long-chain base moieties of 17, 18, and 19 carbons in length, and the 17- or 19-carbon long-chain bases have terminal branches (34). The PE DHC lipids have the same core dihydroceramide structures as the PG DHC lipids but lack the second isobranched fatty acid chain. This structural difference could be critical in producing the differential effects of PG DHC and PE DHC lipid classes. Of note, a recent report has shown that PE DHC lipids stimulate IL-6 secretion from dendritic cells through the engagement of TLR2 (31). Although PE DHC lipids promote impressive activation of dendritic cells in vitro, the lack of PE DHC effects on either WT or TLR2^−/− osteoblasts in vitro suggests that cell activation by PE DHC lipids is in part dependent on the target cell type and culture conditions in addition to the dependence on TLR2 engagement. This differential cell activation of osteoblasts by PE DHC lipids in vitro versus in vivo will be the subject of future investigations.

A recent report demonstrated how various microbial ligands interact with TLR2 heterodimerized with either TLR1 or TLR6 and the dependence of this interaction on the number of fatty acid aliphatic chains within the TLR2 ligand (23). Microbial lipoprotein ligands that contain two aliphatic chains engage TLR2 heterodimerized with TLR6, whereas lipoprotein ligands with three fatty acids, including an amide linked fatty acid, engage TLR2 with TLR1 (23). Our evidence suggests that the three aliphatic chains of the PG DHC lipids are required to demonstrate TLR2-dependent inhibition of osteoblast function in vitro, which may involve coengagement with TLR1. By analogy, PE DHC lipids with two aliphatic lipid chains may coengage TLR2 with TLR6 in dendritic cells. If this is the case, the observed inhibition of osteoblast activity by PG DHC lipids could be related to selective expression of TLR1 over TLR6 by osteoblasts. At this point, there are no reports describing such a selective expression of TLR1 versus TLR6 receptors in osteoblasts or any other cell type. Also possible is that PE DHC lipids may interact with a soluble coreceptor that engages TLR2, and this soluble coreceptor may be present in vivo but not in vitro. Additional TLR2 heterodimer receptors have been reported (36), and one or more of these additional coreceptors could also participate in PG DHC lipid effects on osteoblasts. This question will be the subject of future research.

In summary, the results of the present study indicate a potentially important mechanism for the promotion of alveolar bone loss in humans infected orally with P. gingivalis. Although previous reports suggest that P. gingivalis LPS may be important in promoting periodontal bone loss, two reports have shown that LPS of P. gingivalis is recovered at negligible levels in diseased periodontal tissues in humans (30, 33), whereas the novel complex lipids of P. gingivalis are recovered on periodontally diseased teeth at levels capable of promoting inflammatory responses (32). The recent report demonstrating PG DHC and PE DHC lipid contamination of diseased gingival tissues (35) supports the conclusion of the present investigation that PG DHC and perhaps PE DHC lipids of P. gingivalis are likely to be important in alveolar bone loss in periodontal disease sites. That PG DHC lipids inhibit osteoblasts through engagement of TLR2 is also consistent with TLR2-dependent alveolar bone loss reported using animals infected orally with P. gingivalis (14, 16). The results of the present study provide a new paradigm for bacterial factor alteration of bone cell function with the potential to promote alveolar bone loss in periodontal diseases and perhaps other bone destructive infectious diseases. Furthermore, we have demonstrated here the role of another pathogen-associated molecular pattern (PAMP) that is recognized by TLR2 in modifying osteoblast function.