ABSTRACT
The serine-glycine dipeptide lipid classes, including lipid 430 and lipid 654, are produced by the periodontal pathogen Porphyromonas gingivalis and can be detected in lipid extracts of diseased periodontal tissues and teeth of humans. Both serine-glycine lipid classes were previously shown to engage human and mouse Toll-like receptor 2 (TLR2) and to inhibit mouse osteoblast differentiation and function through engagement of TLR2. It is not clear if other lipids related to serine-glycine lipids are also produced by P. gingivalis. The goal of this investigation was to determine whether P. gingivalis produces additional lipid classes similar to the serine-glycine lipids that possess biological properties. P. gingivalis (ATCC 33277) was grown in broth culture, and lipids were extracted and fractionated by high-performance liquid chromatography (HPLC). Lipids were separated using semipreparative HPLC, and specific lipid classes were identified using liquid chromatography-tandem mass spectrometry (LC-MS/MS) and liquid chromatography-multiple reaction monitoring (LC-MRM) mass spectrometric approaches. Two glycine lipid classes were identified, termed lipid 567 and lipid 342, and these lipid classes are structurally related to the serine-glycine dipeptide lipids. Both glycine lipid classes were shown to promote TLR2-dependent tumor necrosis factor alpha (TNF-α) release from bone marrow macrophages, and both were shown to activate human embryonic kidney (HEK) cells through TLR2 and TLR6 but not TLR1. These results demonstrate that P. gingivalis synthesizes glycine lipids and that these lipids engage TLR2 similarly to the previously reported serine-glycine dipeptide lipids.
INTRODUCTION
Within the complex subgingival microflora of adult periodontitis sites in humans, the periodontal pathogen Porphyromonas gingivalis has been implicated as a critical microbe in chronic inflammatory and alveolar bone destructive processes. Animals expressing defective innate immune receptors have been used to implicate specific host responses to P. gingivalis in mediating alveolar bone loss. For example, oral gavage of wild-type mice with P. gingivalis resulted in significantly reduced alveolar bone levels (greater bone destruction) around posterior teeth when compared with alveolar bone levels in animals not infected with P. gingivalis. However, Toll-like receptor 2 (TLR2) knockout mice did not show reduced alveolar bone levels following oral gavage with P. gingivalis (1–3). Similar results were obtained in another study using P. gingivalis oral gavage combined with ligature placement on teeth (3). These studies indicate that TLR2 engagement contributes to P. gingivalis-mediated alveolar bone loss in mice orally infected with P. gingivalis. By contrast, alveolar bone loss was similar in wild-type and TLR4 knockout mice (defective lipopolysaccharide [LPS] receptor) but was significantly greater than that observed in noninfected mice (2), indicating that P. gingivalis factors other than LPS are critical in promoting alveolar bone loss. Invasion of oral tissues by P. gingivalis, instead of specific cell wall factors penetrating into tissues, is inconsistent with selective TLR2 activation because additional evidence has shown that P. gingivalis exposure to host immune cells will simultaneously engage both TLR2 and TLR4 (4). Therefore, alveolar bone loss following oral gavage with P. gingivalis (1, 2) appears to result from penetration into gingival tissues of P. gingivalis factors that specifically engage TLR2. Complex lipids (5, 6) and lipoprotein (7) of P. gingivalis have been shown to engage human and mouse TLR2, and many novel complex lipid classes of P. gingivalis have been shown to contaminate diseased human periodontal tissues (8). The purpose of this investigation was to characterize two recently identified glycine lipid classes of P. gingivalis and determine if these glycine lipids engage TLR2 as has been observed previously with serine-glycine dipeptide lipids of P. gingivalis.
P. gingivalis produces unusual serine-glycine dipeptide lipids that can engage TLR2 and mediate responses in inflammatory (6) and bone cells (9). The dominant serine dipeptide lipid produced by P. gingivalis is termed lipid 654. Lipid 654 contains a carboxy terminus of serine linked to glycine, and the glycine moiety is amide linked to a beta hydroxy fatty acid, which is usually 3-OH isobranched (iso)-C17:0 (6). The beta hydroxy group can be ester linked to another fatty acid, most frequently to iso-C15:0 (6). P. gingivalis also produces another serine-dipeptide lipid class, called lipid 430, that lacks the esterified fatty acid (6). Lipid 430 is recovered in relatively low levels in P. gingivalis compared with lipid 654. It is not known how these lipids are produced. Hypothetically, serine-glycine dipeptide lipids could be synthesized from the respective glycine lipid precursors by attaching serine to the terminal glycine moiety. The predicted glycine lipid would be either lipid 342 (without an esterified fatty acid) or lipid 567 (with an esterified fatty acid) based on the expected negative ion masses of the dominant molecular species within each class (see Fig. 1H). 3-OH iso-C17:0 is the dominant fatty acid constituent in both serine-glycine dipeptide lipid classes (6), and 3-OH iso-C17:0 is likely to be the dominant fatty acid constituent in glycine lipids. The first goal of this investigation was to evaluate lipid extracts of P. gingivalis for glycine lipids and purify these lipid products using high-performance liquid chromatography (HPLC) and solid-phase extraction (SPE) chromatography. This investigation also sought to evaluate the capacity of these lipid classes to activate mammalian cells and whether TLR2 is engaged in biological responses. Mass spectrometry was used to evaluate lipid extracts for these putative lipid products as has been previously described (6, 10, 11).
Mass spectrometric identification of glycine lipids of P. gingivalis. Negative ion LC-MS was used to characterize preparations of enriched lipid 654 (A), lipid 567 (B), and lipid 342 (C) isolated from P. gingivalis using neutral solvent HPLC followed by subsequent acidic solvent HPLC fractionation of pooled neutral solvent fractions (see Materials and Methods). Treatment of lipid 567 with honey bee venom (HBV) PLA2 and subsequent purification of the hydrolysis products by acidic solvent HPLC recovered a lipid class (D) with negative ion MS/MS mass spectral characteristics identical to that observed with lipid 342 isolated from P. gingivalis (E) and synthetic lipid 342 (F). The distribution of 3-hydroxy (OH) fatty acids in the serine and glycine lipid preparations of P. gingivalis (G) was determined using GC-MS. (H) The distribution of serine-glycine lipids is shown for the three indicated strains of P. gingivalis grown under laboratory conditions as listed in Materials and Methods. (I) Proposed structures are shown for the dominant lipid species of lipid 567 and lipid 342 of P. gingivalis.
Prior work has shown that the lipid 654 class of P. gingivalis and other Bacteroidetes members can be hydrolyzed by the enzyme phospholipase A2 (PLA2), resulting in the lipid 430 class due to release of an esterified fatty acid (12, 13). PLA2 hydrolysis of lipid 654 is stereoselective (12), and the product lipid 430 class engages both mouse and human TLR2 (6). Hydrolysis of lipid 654 by PLA2 was surprising because lipid 654 is not a glycerol-based lipid and does not contain phosphate. Our findings suggested that exposure of lipid 654 to PLA2 in diseased periodontal tissues could result in elevated lipid 430 levels. The question of glycine lipid hydrolysis by PLA2 is also relevant to the present study. Enzymatic hydrolysis of the high-mass glycine lipid class (lipid 567) by PLA2 could result in production of a deesterified glycine lipid class (lipid 342). The second goal of this study was to determine whether PLA2 mediates hydrolysis of the lipid 567 class and whether the product glycine lipid class possesses biological activity.
RESULTS
The first component of this investigation evaluated total lipid extracts of P. gingivalis lipids for the so-called glycine lipids. Single-stage mass spectrometry of the total lipid extract of P. gingivalis (ATCC 33277) revealed that the dominant negative ions for the two proposed glycine lipid classes, termed lipid 567 and lipid 342, are present in P. gingivalis (data not shown). Next, P. gingivalis total lipids were sequentially fractionated first with neutral solvent HPLC fractionation followed by pooling of appropriate fractions by lipid class. The pooled fractions were then separated by acidic solvent HPLC fractionation. With this approach, highly enriched preparations of lipid 654, lipid 567, or lipid 342 were obtained. Lipid 430 was recovered in very low amounts and was not included in the comparisons. Mass spectra of the glycine lipids of P. gingivalis are shown in Fig. 1. Three characteristic lipid species within lipid 567 (Fig. 1B) and lipid 342 (Fig. 1C), all varying by methyl group increments (14 atomic mass units [amu]), are evident in these mass spectra, consistent with the characteristic three species observed in lipid 654 (Fig. 1A) as previously reported (6). These lipid species represent methyl group deletions in the fatty acid constituents as described below. Figure 1D also shows that lipid 567 is converted to lipid 342 through the action of honey bee venom phospholipase A2 as has been reported for lipid 654 deesterification to lipid 430 by PLA2 (13). Figure 1E also shows the characteristic tandem mass spectrometry (MS/MS) spectra of the dominant lipid species of the lipid 342 class recovered from P. gingivalis (m/z 342.2). Synthetic standard of lipid 342 showed essentially the same MS/MS spectrum (Fig. 1F), thus confirming the predicted structure of the dominant lipid species of the lipid 342 class of P. gingivalis. MS/MS spectra of lipid 567 revealed only one major collision-induced ion (m/z 324) for both the synthetic standard and the dominant lipid species of lipid 567 isolated from P. gingivalis (data not shown). MS/MS analysis of synthetic standards of the dominant lipid species of lipid 567 and lipid 342, therefore, demonstrated identical MS/MS spectra for the dominant lipid species of lipid 567 or lipid 342 isolated from P. gingivalis. Finally, aliquots of lipid 654, lipid 567, and lipid 342 were hydrolyzed under alkaline conditions (4 N KOH, 2 h, 100°C) and derivatized to form pentafluorobenzyl ester, trimethylsilyl ether derivatives. The resultant fatty acid products, analyzed by gas chromatography-mass spectrometry (GC-MS), revealed that 3-OH iso-C17:0 is the dominant hydroxy fatty acid recovered in lipids 654, 567, and 342 (Fig. 1G). 3-OH C16:0 and 3-OH iso-C15:0 were considerably less abundant. Note that the abundance of the three 3-OH fatty acids parallels the abundances of the three characteristic lipid species within the lipid 654, lipid 567, and lipid 342 classes (Fig. 1A to C). Since 3-OH C19:0, 3-OH C18:0, 3-OH C14:0, and 3-OH C13:0 are negligible in these hydrolyzed lipid samples (data not shown), we conclude that the order of fatty acid substitution in decreasing abundance within lipid 342 and lipid 567 species is 3-OH iso-C17:0, 3-OH C16:0, and 3-OH iso-C15:0, all being amide linked to glycine. This analysis confirms the structures of lipid 342 and lipid 567 species in P. gingivalis.
Using the developed liquid chromatography-multiple reaction monitoring (LC-MRM) method (see Materials and Methods), the relative distributions of glycine and serine lipid classes were determined in total lipid extracts from three strains of P. gingivalis and revealed that lipid 567 is the dominant lipid class (Fig. 1H). Because lipid 567 is clearly more abundant than other serine-glycine lipids, the biological properties of this lipid class along with the related lipid 342 class were evaluated. In order to evaluate biological properties of these glycine lipid classes, multiple batches of P. gingivalis total lipids (B17/18 and B19) were fractionated by neutral solvent semipreparative HPLC as described in Materials and Methods. Fractions were pooled based on the recovery of lipid 654, lipid 567, or lipid 342. These pooled lipid fractions were then subjected to acidic HPLC fractionation, after which the eluted lipid fractions were evaluated by LC-MRM for the relative levels of serine and glycine lipid classes. The relative enrichment for each lipid product, based on LC-MRM quantitation of characteristic ion transitions, is shown for specific lipid preparation batches and their respective acidic solvent HPLC fractions (Fig. 2A). Lipid 654 preparations were contaminated to some extent with lipid 567 or lipid 342, but highly enriched preparations of lipid 567 and lipid 342 were obtained (Fig. 2A). The highly enriched preparations of the serine or glycine lipids, including lipid 654, lipid 567, and lipid 342, consistently stimulated tumor necrosis factor alpha (TNF-α) release from bone marrow macrophages in a TLR2-dependent manner (Fig. 2B).
HPLC enrichment of lipid 654, lipid 567, and lipid 342. Total lipids were extracted from P. gingivalis (ATCC 33277) and are indicated by the batch number of the extracted lipids. Fractions from batches 17 and 18 were pooled. Serine-glycine lipids were enriched using neutral solvent HPLC, and specific fractions were pooled as indicated (Pooled Neutral Fr). The pooled fractions were refractionated with acidic solvent HPLC. Acidic fraction (Afr) numbers are indicated. (A) The relative enrichment for each serine-glycine lipid class is indicated within each acidic HPLC fraction. (B) The capacity of each fraction to promote TNF-α release from bone marrow macrophages is shown. Bone marrow macrophages were exposed to each lipid preparation at a concentration of 1 μg/ml for 24 h followed by collection of medium samples for detection of TNF-α by ELISA. Each histogram bar represents the average of two replicate determinations from a single culture well of murine bone marrow macrophages, and error bars represent the standard deviation of the mean. KO, knockout.
The specificity of lipid 654 and lipid 567 preparations to promote TNF-α release was further evaluated by applying a solid-phase extraction (SPE) preparative chromatography to additional preparations of lipid 654 and lipid 567. When applied to additional batches (B16 or pooled lipid 654 batches) of acidic purified serine and glycine lipids, the SPE method reduced neutral lipid contamination of the recovered lipid 567 or lipid 654 preparations as shown in Fig. 3. Neutral lipids were first removed using chloroform elution (PA) over Isolute silica gel SPE columns (Biotage LLC, Charlotte, NC). Next, methanol in acetone (2:8, vol/vol) was used to elute predominantly lipid 567 and lipid 654 (PB). Methanol was then used to elute primarily lipid 654 (PC) as determined using LC-MS (Fig. 3A). Acidic fractions 14 to 16 of the batch 16 lipid 654 preparation contained varying amounts of neutral lipid that eluted in chloroform (PA). Also, note that the pooled lipid 654 sample from early batches of P. gingivalis total lipids contained a considerable amount of lipid 567 in solvent fraction PB, similar to the contamination of lipid 654 shown in Fig. 2A. All SPE fractions were then screened for promotion of TLR2-dependent TNF-α release from mouse bone marrow macrophages (Fig. 3B). Bone marrow macrophages from wild-type mice that were not treated with bacterial lipids (negative controls) released less than 20 pg/ml of TNF-α. Fractions highly enriched for either lipid 567 or lipid 654 consistently promoted TNF-α release (Fig. 3B), whereas neutral lipids recovered in fraction PA demonstrated no capacity to stimulate TNF-α release.
Solid-phase extraction (SPE) of lipid 654 preparations. Total lipids extracted from P. gingivalis (batch 16) were fractionated using neutral solvent HPLC. Late eluting fractions were pooled and refractionated using acidic solvent HPLC. Acidic fractions 14 to 16 contained the serine-glycine dipeptide lipids and glycine lipids. Lipid 654 was prepared from multiple early batches of P. gingivalis total lipids (ATCC 33277) using the same neutral solvent HPLC followed by acidic solvent HPLC. The indicated acidic HPLC fractions from Batch 16 P. gingivalis lipids or the pooled Lipid 654 preparation were then fractionated using an SPE technique described in Materials and Methods. The SPE technique included sequential elution over a C18 reverse phase column, first with 4 ml of chloroform (PA), next with 5 ml of methanol/acetone (8:2, vol/vol, PB), and finally with 5 ml of methanol (PC). (A) The total lipid (mg) recovered in each SPE fraction is listed above each histogram bar, and the distribution of serine-glycine lipids within the respective SPE solvent phases is also shown. (B) The capacity of each fraction to promote TNF-α release from wild-type or TLR2 knockout bone marrow macrophages is shown. Bone marrow macrophages were exposed to each lipid preparation at a concentration of 1 μg/ml for 24 h followed by collection of medium samples for detection of TNF-α by ELISA. Each histogram bar (B) represents the average of two replicate determinations from a single culture well of murine bone marrow macrophages, and error bars represent the standard deviation of the mean. TNF-α release is depicted as the mean ± standard deviation (n = 2 determinations) for each lipid preparation.
Highly purified lipid 654, lipid 567, and lipid 342 prepared from another batch of P. gingivalis (33277) and synthetic lipid 342 were then evaluated for engagement of human TLR2 and related coreceptors using human embryonic kidney (HEK) cells transfected with specific Toll-like receptors (TLRs). Engagement of the indicated Toll-like receptors leads to activation and translocation of NF-κB together with expression of the secretory alkaline phosphatase (SEAP) reporter gene and subsequent release of SEAP from the transfected HEK cells. Extracellular SEAP is then detected in culture medium using a commercially available medium containing the alkaline phosphatase substrate (InvivoGen). Toll-like reeceptor activation in HEK cells is easily quantified using SEAP expression, although we have also used interleukin-8 (IL-8) secretion from HEK cells as another surrogate measure of TLR activation. TNF-α is not produced to a significant extent by activated HEK cells and, therefore, cannot be compared with TNF-α secretory responses from bone marrow macrophages.
The relative purity of each lipid preparation isolated from P. gingivalis is shown in Table 1. Figure 4A shows the effects of low-dose lipid 342, synthetic lipid 342, and lipid 654 on HEK NF-κB secretory alkaline phosphatase (SEAP) reporter cells transfected with human TLR2. Lipid 342 produced by hydrolysis of lipid 567 with PLA2 (Hyd L342, 1 μg/ml) significantly increased SEAP expression over control cultures, and all doses of synthetic lipid 342 and lipid 654 significantly increased SEAP expression. By contrast, TLR4-transfected HEK cells showed little stimulation by P. gingivalis serine or glycine lipids or the Pam2Cys (P2C) (Fig. 4B). Although lipid 567 at a dose of 0.1 μg/ml showed a significant increase in TLR2 engagement, this was not consistent with the high or low dose of this lipid class and was concluded to be an outlier result for unknown reasons. However, Salmonella enterica serovar Typhimurium LPS (0.1 μg/ml) significantly increased SEAP expression in TLR4-transfected HEK cells, consistent with known engagement of TLR4 by LPS.
Lipid preparations used for testing HEK cellsa
Serine-glycine dipeptide lipid and glycine lipid stimulation of human embryonic kidney (HEK) cells transfected with TLR2 (A) or HEK cells transfected with TLR4 (B). HEK cells were cultured according to the supplier’s instructions. HEK TLR2 cells were transfected with human TLR2 and CD14, whereas HEK TLR4 cells were transfected with human TLR4, MD-2, and CD14. Cells were suspended in development medium for secretory alkaline phosphatase (SEAP), and the mixture was added to assay wells previously supplemented with sonicated lipid preparations at the indicated doses. HEK cells were treated with lipids for 16 h, at which time the multiwell culture dishes were read at 630 nm for secretory alkaline phosphatase (SEAP) activity. The lipopeptide Pam2Cys (P2C) and Salmonella Typhimurium lipopolysaccharide (St LPS) served as positive controls for TLR2 and TLR4, respectively. Each histogram bar represents the mean ± standard deviation for n = 4 replicate determinations for one representative experiment. One-factor ANOVA with pairwise comparisons using Dunnett’s multiple-comparison test revealed significant differences versus the control cultures (#, P < 0.05).
Bacterial lipid effects were compared between TLR2-transfected HEK cells and Null1 HEK cells. Lipid 654 (1 μg/ml) and P. gingivalis total lipids (1 μg/ml) significantly stimulated SEAP release from TLR2-transfected HEK cells but produced no effect in Null1 HEK cells (Fig. 5). Low-dose serine or glycine lipids were comparable to the total lipid extract of P. gingivalis in engaging TLR2. The positive lipopeptide controls Pam3Cys and Pam2Cys (10 ng/ml each) gave strong SEAP responses. Note that the dose of neutralizing antibody for human TLR2 (10 μg/ml) did not suppress TLR2 engagement by Pam2Cys at 10 ng/ml and only partially suppressed TLR2 engagement with P3C (Fig. 5B). However, an additional experiment exposed HEK TLR2 cells to Pam2Cys (100 pg/ml) plus either anti-human TLR2 antibody or isotype control antibody (10 μg/ml each). This trial revealed that the lower dose of Pam2Cys significantly stimulated SEAP release in the presence of the isotype control antibody, but SEAP release was virtually completely attenuated by anti-human TLR2 antibody treatment (see Fig. 5B, rightmost histogram bars). We conclude that anti-human TLR2 antibody at 10 μg/ml cannot overcome the stimulatory effect of a high dose of Pam2Cys (10 ng/ml) in the culture medium. SEAP release from HEK TLR2 cells treated with neutralizing antibody or isotype control antibody revealed that synthetic lipid 342, lipid 567, and lipid 654 effects were inhibited specifically by anti-human TLR2 neutralizing antibody treatment but not by the isotype control antibody (Fig. 5B).
Serine-glycine dipeptide lipid and glycine lipid stimulation of human embryonic kidney (HEK) cells transfected with TLR2 versus Null1 HEK cells (A) or TLR2 transfected HEK cells treated with anti-human TLR2 neutralizing antibody (anti hTLR2) versus antibody isotype control (B). HEK cells were cultured according to the supplier’s instructions. Cells were suspended in development medium for secretory alkaline phosphatase (SEAP) and added to 96-well culture dishes with specific wells supplemented with sonicated lipid preparations at the indicated doses. Antibody preparations (10 μg/ml) were added to culture wells along with the lipid preparations. HEK cells were treated with lipids for 22 h, at which time the multiwell culture dishes were read at 630 nm for SEAP activity. Each histogram bar represents the mean ± standard deviation for n = 4 replicate determinations for one representative experiment. The effect of anti-human TLR2 neutralizing antibody on Pam2Cys (P2C, 0.1 ng/ml) was evaluated in a separate experiment. One-factor ANOVA with pairwise comparisons using Dunnett’s multiple-comparison test revealed significant differences versus the control cultures (# and &, P < 0.05).
In contrast to the singly transfected cells, SEAP release from TLR2/TLR6-transfected HEK cells was significantly elevated over controls for all bacterial lipid preparations and doses (Fig. 6A and B). SEAP release from TLR2/TLR1-transfected HEK cells was either not affected by the indicated lipid preparations or was only slightly increased. Note that the Pam3Cys stimulation of HEK cells was observed only with TLR2/TLR1-transfected cells, whereas Pam2Cys stimulation of HEK cells was observed only with TLR2/TLR6-transfected cells. Anti-human TLR6 antibody (10 μg/ml) attenuated SEAP release from TLR2/TLR6-transfected HEK cells treated with the indicated bacterial lipids but was not inhibited by the isotype antibody control (Fig. 6C). Pam2Cys stimulation of SEAP release was also partially inhibited by anti-human TLR6 antibody (see Pam2Cys dose effects described above). In addition, HEK cells transfected with both TLR2 and TLR6 showed greater activation than HEK cells transfected with only human TLR2. HEK cells transfected with TLR2/TLR1 did not respond to lipid 342 or lipid 654 but did respond to lipid 567 in a dose-dependent manner. This could indicate that the TLR1 receptor can engage the P. gingivalis lipid 567 structures. Lipid 567 and lipid 654 demonstrated essentially equivalent potencies in engaging TLR2/TLR6, which indicates that the terminal serine moiety does not enhance TLR2-mediated effects. Specificity of lipid effects on human TLR2 or human TLR6 was demonstrated using the respective anti-human neutralizing antibody for each of these receptors. These results showed that both P. gingivalis lipid 342 and synthetic lipid 342 engage human TLR2/TLR6 though this lipid class contains only a single acyl chain. Furthermore, transfection of HEK cells with TLR2/TLR6 increased the magnitude of bacterial lipid effects on SEAP release when compared with bacterial lipid effects on HEK cells transfected with only TLR2.
Serine-glycine dipeptide lipid and glycine lipid stimulation of human embryonic kidney (HEK) cells transfected with TLR2/6 (A) versus HEK cells transfected with TLR2/TLR1 (B) and TLR2/TLR6-transfected HEK cells treated with anti-human TLR6-neutralizing antibody versus antibody isotype control (C). According to the commercial supplier, endogenous genes for TLR1 and TLR6 were neutralized by double knockout before transfection with either human TLR2/TLR6 or human TLR2/TLR1. Both cell lines were also transfected with human CD14. HEK cells were cultured according to the supplier’s instructions. Cells were suspended in development medium for secretory alkaline phosphatase (SEAP) and added to 96-well culture wells supplemented with sonicated lipid preparations at the indicated doses. Antibody preparations (10 μg/ml) were added to culture wells along with the lipid preparations. HEK cells were treated with lipids for 26 h, at which time the culture dishes were read at 630 nm for SEAP activity. Each histogram bar represents the mean ± standard deviation for n = 4 replicate determinations for one representative experiment. One-factor ANOVA with pairwise comparisons using Dunnett’s multiple-comparison test revealed significant differences versus the respective control cultures (# and &, P < 0.05).
DISCUSSION
Mass spectrometric analysis of the total lipids of P. gingivalis revealed not only serine-glycine dipeptide lipids but also the presence of glycine lipid classes, termed lipid 567 and lipid 342, based on the negative ion masses of the respective dominant lipid species in each class. Of the four lipid classes produced, lipid 567 was consistently the most abundant within the total lipid extracts from three strains of P. gingivalis grown in broth culture (Fig. 1H). The method used to detect and verify identity of the serine-glycine dipeptide lipid and glycine lipids included both reverse-phase ultraperformance liquid chromatography (UPLC) verification of lipid retention times and mass spectrometric collision-induced dissociation of lipid classes for identification and quantitation of ion fragments (multiple reaction monitoring [MRM]). Using this analytical approach, lipid 342 and lipid 567 recovered from the three strains of P. gingivalis (Fig. 1H) were observed to be structurally identical, and it was concluded that isolation of these lipid classes from each strain of P. gingivalis was not essential for biological testing purposes. Therefore, we isolated lipid 654, lipid 567, and lipid 342 only from the 33277 strain of P. gingivalis. The structure of lipid 567 was originally described in lipid extracts of Cytophaga johnsonae and was termed cytolipin (14). Lipid 342-like structures were also previously described as minor lipid constituents of specific organisms of the phylum Bacteroidetes. The primary glycine lipid product previously identified in the human gut microbiome was termed commendamide with a structure of N-acyl-3-hydroxypalmitoyl-glycine (15), which differs from lipid 342 by a single methyl group deletion at the terminal end of the fatty acid aliphatic chain. More recently, glycine lipids recovered from Bacteroides thetaiotaomicron, a common Bacteroides found in the mouse and human gastrointestinal microbiome, were proposed to include a diacylated glycine lipid containing either 3-OH-methyl 16:0+15:0 or 3-OH-methyl 15:0+16:0 (both with a negative ion mass of m/z 566.4) and a monoacylated glycine lipid containing a single acyl chain of either 3-OH 17:0 or 3-OH-methyl 16:0 (both with a negation ion mass of m/z 342.2) (16). Based on the fatty acid analysis of lipid 567 isolated from P. gingivalis, the dominant hydroxy fatty acid recovered was 3-OH iso-C17:0 (3-OH-methyl C16:0). Like lipid 342 isolated from P. gingivalis, the dominant hydroxy fatty acid recovered in lipid 567-derived lipid 342 was 3-OH iso-C17:0, consistent with the acyl chain substitutions previously described for lipid 654 and lipid 430 (6). Based on the hydroxy fatty acids recovered in lipid 342 and lipid 567 hydrolyzates, the lower abundance m/z 328 and 314 negative ions within the lipid 342 class (see Fig. 1C and D) reflect the recovery of glycine lipids containing 3-OH C16:0 and 3-OH iso-C15:0, respectively. Structural confirmation of lipid 342 was demonstrated by showing identical retention times and mass spectra of trimethylsilyl (TMS) derivatives of synthetic lipid 342, the dominant lipid species recovered in P. gingivalis lipid 342 and lipid 342 derived from hydrolyzed lipid 567 (data not shown). As with lipid 567, the dominant lipid species in lipid 342 from P. gingivalis contains 3-OH iso-C17:0 amide linked to glycine, and the other two less abundant species (m/z 328 and 314) contain 3-OH C16:0 and 3-OH iso-C15:0, respectively.
Neutral solvent HPLC fractionation of P. gingivalis total lipids followed by mass spectrometric screening revealed lipid fractions enriched in lipid 654, lipid 567, or lipid 342. When the neutral solvent HPLC fractions were pooled for lipid 654, lipid 567, or lipid 342 and subsequently separated using acidic solvent HPLC, highly enriched preparations of lipid 654, lipid 567, and lipid 342 were recovered in early eluting fractions (Fig. 2A). Each of these enriched fractions of glycine or serine lipids of P. gingivalis promoted TNF-α release from bone marrow macrophages and required TLR2 engagement for cell activation. In order to exclude nonpolar neutral lipids as possible cell-activating lipids, a solid-phase extraction (SPE) method provided further enrichment of contaminated lipid 654 and lipid 567 preparations (Fig. 3A). Nonpolar neutral lipids eluted in the first solvent fraction (100% chloroform indicated as PA). Neutral lipids typically do not ionize under liquid chromatography-mass spectrometric analysis, making their identification difficult. Nevertheless, these neutral lipids can be contaminants of the lipid 567 or lipid 654 isolates. Figure 3 shows that lipid 654 and lipid 567 do not elute with neutral lipids recovered in the first elution fraction (SPE solvent PA), and the neutral lipids do not promote TNF-α release from macrophages. Only fractions containing lipid 567, lipid 654, and/or lipid 342 promoted TNF-α release from bone marrow macrophages (Fig. 3B).
Based on the results using murine bone marrow macrophages and HEK cells transfected with specific Toll-like receptors, lipid 342 of P. gingivalis is a monacyl agonist of TLR2. As has been previously reported for lipid 430, the single acyl chain of lipid 342 allows engagement of TLR2, although lipopeptides containing either two or three acyl chains promote TLR2-mediated responses at lower doses (17). Commercially available Pam2Cys and Pam3Cys lipopeptides contain either two or three palmitic acid acyl chains, respectively, and this fatty acid is not branched. Of note, the lipopeptide prepared by trypsin hydrolysis of a native lipoprotein of P. gingivalis demonstrated two palmitic acids (C16:0) and one pentadecanoic acid (C15:0) in the terminal glycerocysteine moiety of the lipopeptide (7). As reported subsequently, synthetic standards of the P. gingivalis lipopeptide, including the three positional fatty acid isomers of the triacylated glycerocysteine, were shown to engage murine TLR2 regardless of the order of fatty acid substitutions on glycerocysteine moiety (18). However, TLR1 and TLR6 were not engaged by these lipopeptide preparations regardless of the unbranched fatty acid substitutions on the glycerocysteine (18). By contrast, lipid 567 and lipid 654 of P. gingivalis engage both human TLR2 and TLR6 but not TLR1 and are substituted predominantly with branched chain 3-OH iso-C17:0 and iso-C15:0. The activation of HEK cells by lipid 342, lipid 567, and lipid 654 suggests that the isobranched aliphatic chains of P. gingivalis glycine or serine lipids are contributing to the specific engagement of TLR2 and TLR6. By contrast, the LPS of P. gingivalis contains isobranched fatty acids within its lipid A component, including 3-OH iso-C17:0 (19), but this LPS does not significantly engage TLR2 at a concentration of 0.1 μg/ml (20). The non-fatty acid components of P. gingivalis LPS or its associated lipid A are, therefore, likely to prevent engagement of TLR2.
We have previously reported that lipid 654 of P. gingivalis contains virtually exclusively l-serine (9) and the beta carbon in 3-OH iso-C17:0 is observed only in the (R) epimeric form in lipid 654 when isolated from P. gingivalis, subgingival calculus, and other oral bacteria of the phylum Bacteroidetes (12). Other work has determined that the (R) diastereomeric form of lipid 654, but not the (S) diastereomeric form, can be hydrolyzed by mammalian PLA2 enzyme isoforms (12). We report here that lipid 567 is also hydrolyzed to lipid 342 by honey bee venom PLA2, suggesting that lipid 567 of P. gingivalis exists primarily in the (R) stereoisomeric form. The previously reported hydrolysis of lipid 654 (12) together with the observed hydrolysis of lipid 567 shown here suggests that there is the potential that elevated PLA2 enzymatic activity associated with chronic periodontitis tissues could be associated with increased hydrolysis of lipid 567 to lipid 342 in diseased tissues once lipid 567 gains entry to the diseased tissues. Elevated PLA2 activity in diseased periodontal tissues is inferred because of the known increased levels of prostaglandins demonstrated as periodontal disease severity worsens (21–23). We are currently evaluating the capacity of chronic inflammatory cells to hydrolyze lipid 654 and 567 to lipid 342.
The present study demonstrated that glycine lipids of P. gingivalis engage both human and mouse TLR2. Murine bone marrow macrophages were derived from either wild-type or TLR2 knockout animals, and the enriched serine-glycine lipids of P. gingivalis increased TNF-α release only from wild-type macrophages but not from TLR2 knockout macrophages. Human embryonic kidney (HEK) cells transfected with TLR2 or TLR2/TLR6 were also activated when exposed to lipid 342 or lipid 567, and neutralizing antibody for either TLR2 or TLR6 attenuated the effects of both lipid classes. The isotype antibody controls for each neutralizing antibody did not attenuate the engagement of TLR2 alone or TLR2/TLR6 by either lipid 342 or lipid 567. The HEK cell results indicate that glycine lipid classes of P. gingivalis activate human cells through engagement of TLR2 and TLR6 and that the expression of both receptors is necessary for maximal effects of these lipids.
The present study demonstrates that P. gingivalis produces novel glycine lipids and showed that these lipids activate mouse bone marrow macrophages and human embryonic kidney cells through engagement of TLR2. In the case of lipid 342, engagement of TLR2 requires only a single glycine amino acid that is amide linked to a single hydroxy fatty acid with 3-OH iso-C17:0 being the dominant fatty acid. Glycine lipids are presumed to represent precursor lipids in the synthesis of serine-glycine dipeptide lipids in P. gingivalis but future studies will evaluate these processes in greater detail. Because lipid 654 conversion to lipid 430 and lipid 567 conversion to lipid 342 are mediated by PLA2, it is important to evaluate not only the uptake of these lipids into cells exposed to bacteria but also the capacity of mammalian cells to convert the parent lipid 654 and lipid 567 to smaller hydrolytic products, particularly lipid 342. We have previously shown that lipid 430 is increased in human artery walls, particularly where atheroma formation is evident (13). Because lipid 654 and lipid 430 are recovered in human arteries and diseased periodontal tissues (6), it is important to understand the processes underlying the transport of these lipids into tissues and to determine which cells are capable of mediating hydrolysis of lipid 654 and lipid 567 to lipid 342. Further work is required to determine how these processes are related to tissue destruction associated with chronic inflammatory disease at sites where these lipids are known to accumulate.
MATERIALS AND METHODS
Reagents and cell lines.Porphyromonas gingivalis strains 33277, W83, and 381 were obtained from ATCC (Manassas, VA). Pentafluorobenzyl bromide, silylation-grade acetonitrile, and diisopropyl triethylamine were obtained from Thermo Scientific (Waltham, MA), and bis-trimethylsilyl-trifluoroacetamide (BSTFA) was obtained from Sigma-Aldrich Co. (St. Louis, MO). Cell media were obtained from Gibco (Grand Island, NY). HPLC columns were obtained from Supelco (Sigma-Aldrich Supelco, Inc., Bellefonte, PA), and HPLC solvents and other organic solvents were obtained from Fisher Scientific (Waltham, MA).
Bacterial lipid extraction.P. gingivalis (ATCC 33277, type strain) was grown in broth medium containing basal (peptone, Trypticase, brain heart infusion, and yeast extract [BBL; Fisher Scientific]) medium supplemented with hemin and menadione (Sigma-Aldrich) (10). The suspension cultures were incubated in an anaerobic chamber flushed with N2 (80%), CO2 (10%), and H2 (10%) at 37°C for 4 to 5 days, and the bacteria were harvested by centrifugation (3,000 × g for 20 min). P. gingivalis W83 and 381 strains were extracted using the same method. Following lyophilization, approximately 4 g of P. gingivalis (ATCC 33277) pellet was extracted overnight using a modification of the phospholipid extraction procedure of Bligh and Dyer (24) and Garbus et al. (25). Specifically, 4 ml of H2O + 16 ml of MeOH:CHCl3 (2:1, vol/vol) was added to the bacterial sample and vortexed. After 12 h, 6 ml of 2 N KCl + 0.5 M K2HPO4 and 6 ml CHCl3 were added and the sample vortexed. The lower organic phase was carefully removed, and CHCl3 (6 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.
HPLC methods.Fractionation of bacterial lipids by high-performance liquid chromatography (HPLC) was accomplished using a semipreparative HPLC column (1× 25 cm silica gel, Ascentis, 5 μm; Sigma-Aldrich Supelco, Inc., Bellefonte, PA) and eluted isocratically with hexane/isopropanol/water (6:8:0.75, vol/vol/vol; solvent A) (10, 11). Lipid samples were dissolved in solvent A to achieve a concentration of approximately 14 mg/ml. The sample was centrifuged at 2,500 × g for 10 min and the supernatant removed for HPLC fractionation. Semipreparative HPLC fractionation was accomplished using an HPLC system equipped with dual pumps (LC-10ADvp), automated controller (SCL-10Avp), and in-line UV detector (SPD-10Avp) (Shimadzu Scientific Instruments, Kyoto, Japan). For each chromatographic separation, 7 to 10 mg of lipid was applied, and fractions were pooled for 40 column fractionations. Samples were eluted at 1.4 ml/min and monitored at 205 nm with 2 min fractions collected. Fractions were dried under nitrogen and resuspended in 2 ml of CHCl3. Lipid recovery in each HPLC fraction was determined by drying 5 μl from each fraction and weighing the sample using a Cahn Electrobalance. An additional 10 μl from each fraction was analyzed for complex lipid constituents using liquid chromatography-electrospray ionization-multiple reaction monitoring (LC-ESI-MRM) to determine which HPLC fractions contained lipid 654, lipid 567, lipid 430, and lipid 342 as well as phosphorylated dihydroceramide lipids. Fractions containing specific serine, glycine, and dihydroceramide lipids were pooled and repurified using the acidic fractionation described below.
Fractions containing glycine or serine lipid classes, including lipid 654, lipid 567, and lipid 342, were further purified by elution over the same type of HPLC column at 1.0 ml/min but using HPLC solvent A supplemented with 0.1% acetic acid. Purity of lipid 567 and lipid 342 was verified by electrospray-mass spectrometry (ESI-MS or MS/MS) as described for specific lipid preparations in Results. Following hydrolysis of lipid 567 with honey bee venom PLA2 for 5 days (100 U [Sigma-Aldrich] added to 20 μg of lipid 567 in 10 mM Tris, pH 7.5, 10 mM CaCL2, 150 mM NaCl, 37°C), separation of lipid 567 from lipid 342 was accomplished using the same acidic fractionation described above. Lipid 567 and lipid 342 mass spectral characteristics were verified using LC-MS and LC-MS/MS using a Waters Acquity UPLC coupled with a TQD tandem mass spectrometer (Waters Co., Milford, MA). Quantification of the target lipid classes was accomplished using electronic integration of lipid ion transition peaks.
Synthesis of 13-methyltetradecanoic acid and 3-hydroxyl-15-methylhexadecanoic acid.In Fig. 7, the synthesis of lipid 342 began with the synthesis of the C17 amino alcohol (compound 5). The synthesis used the identical route to that reported for the synthesis of lipid 430 and lipid 654 (26). The Grignard reagent prepared from 1-bromo-2-methylpropane was coupled to 11-bromoundecan-1-ol (compound 1) in the presence of a lithium tetrachlorocuprate complex with N-methylpyrrolidone as an additive (27) to give 13-methyltetradecan-1-ol (compound 2). Subsequent oxidation with 4-acetamido-2,2,6,6-tetramethylpiperidine 1-oxyl) on silica in dichloromethane (28–30) gave 13-methyltetradecanal (compound 3) in 91% yield from compound 1. This aldehyde was oxidized to 13-methyltetradecanoic acid (compound 4) upon exposure to air. The reaction of aldehyde (compound 3) with ethyl diazoacetate in the presence of catalytic tin (II) chloride (27, 31) gave β-hydroxy ester (compound 5) in 84% yield. Reduction with sodium borohydride gave racemic β-hydroxy ester (compound 6), and saponification using lithium hydroxide yielded racemic β-hydroxy acid (compound 5) in an overall yield of 62% from compound 1.
Synthesis of 13-methyltetradecanoic acid (compound 4) and 3-hydroxyl-15-methylhexadecanoic acid (compound 6).
Synthesis of lipid 342 and lipid 567.In Fig. 8, the methyl ester of glycine was prepared by the reaction of glycine with thionyl chloride in methanol. The reaction of glycine methyl ester with compound 6 in the presence of (3-dimethylaminopropyl)-N’-ethylcarbodiimide hydrochloride (EDC-HCl; Sigma-Aldrich) and a catalytic amount of N,N-dimethylaminopyridine (DMAP) in dichloromethane gave a 90% yield of glycine C17-methyl ester (compound 7). Saponification with LiOH gave the free acid (compound 8) (lipid 342) in greater than 70% yield. The preparation of lipid 567 was accomplished by the reaction of 13-methyltetradecanoic acid (compound 4) with compound 7 in the presence of EDC-HCl and a catalytic amount of DMAP in dichloromethane.
Synthesis of lipid 342 (compound 7) and lipid 567 (compound 8).
Bacterial fatty acid analysis.Analysis of fatty acids within each serine-glycine lipid preparation was accomplished by treating a sample of each lipid fraction with 4 N KOH (0.5 ml, 100°C, 2 h). Hydrolysates were cooled, and 200 μl of HCl and 1 ml of water were added to each sample. Each sample was then extracted with chloroform (3× 2 ml) and the pooled extracts dried under nitrogen. Fatty acids were treated with pentafluorobenzyl (PFB) bromide/diisopropylethylamine/acetonitrile (6.6:20:73.3, vol/vol/vol, 50 μl, 20 min, 50°C) and dried under nitrogen. The samples were then treated with (bis)trimethylsilyl (TMS)-trifluoroacetamide (40 μl, 50°C, 1 h) and transferred to autosampler vials for GC-MS analysis. Synthetic standards of isobranched fatty acids, saturated fatty acids, and 3-OH fatty acids obtained from Matreya, Inc. (Pleasant Gap, PA) were treated to form PFB-TMS derivatives using the same method.
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 by 0.25 mm, 0.25-μm film thickness; Agilent Technologies, Lexington, MA) 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). The injector block was held at 285°C, and the transfer tube was maintained at 290°C. Mass spectra were acquired using negative ion chemical ionization mode, and bacterial fatty acid products were identified by retention time of characteristic negative ion peaks (molecular ions – H). Fatty acid identities were verified against synthetic fatty acid standards.
Lipid preparation for cell treatment.Predetermined amounts of enriched serine or glycine lipid fractions were transferred in acidic HPLC solvent to conical vials and dried under nitrogen. The selected HPLC fractions were then supplemented with cell culture medium to achieve the indicated final lipid concentration. The lipids were then sonicated for 15 s at 3 W, after which the stock lipid preparations were added to cell culture medium. Lipids were added to medium at a dose of 1 μg/ml or less. For the lipid classes under evaluation, a single carboxylic acid group is present in each lipid class, and using the predicted lipid masses, the amount of acidic lipid added to each culture would be less than 10 nmol for all classes. This is an insufficient amount of acidic substance to alter the pH of the culture medium. Other vehicles to increase solubility of the lipids in aqueous culture medium were not used in this study.
Murine cell culture.Bone marrow macrophages were cultured from bone marrow cells collected from femurs of 4- to 8-week-old wild-type or TLR2 knockout C57BL mice. Mice were sacrificed and femurs recovered in accordance with a protocol approved by the University of Connecticut IACUC. The ends of the long bone shafts were cut, and the marrow cells were centrifuged from the marrow spaces. The cells were resuspended, counted, and dispersed in culture dishes in Dulbecco modified Eagle medium (DMEM) containing 100 U/ml penicillin, 100 μg/ml streptomycin, 250 ng/ml amphotericin B (Gibco), and 10% fetal calf serum. After overnight culture, nonadherent cells were collected by pipetting the culture dish and the cells centrifuged. Cells were dispersed at a density of 10 × 106 cells/15 ml in 100- by 20-mm petri dishes and treated for 6 days with medium containing 15% L929 fibroblast-conditioned medium. Medium was replaced at 3 days. At 6 days, the adherent cells were recovered by pipetting, counted, and plated into multiwell dishes at a density of 0.15 × 106 cells/cm2. The cells were then treated with the sonicated bacterial lipid preparations. After 24 h of lipid exposure, medium was harvested and frozen.
ELISA for TNF-α.Previously collected medium samples were thawed and analyzed for TNF-α without dilution. TNF-α was quantified using an enzyme-linked immunosorbent assay (ELISA) for murine TNF-α (R&D Systems, Minneapolis, MN).
HEK cell culture.HEK cells (InvivoGen, San Diego, CA) were used to evaluate human Toll-like receptor engagement by P. gingivalis lipids. TLR2-, TLR4-, TLR2/TLR6-, and TLR2/TLR1-transfected HEK cell lines as well as Null1 HEK cells were grown in DMEM 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 [InvivoGen] and S. Typhimurium LPS [Gibco]) were added to 96-well culture dishes to achieve the indicated concentrations. Controls included treatment of HEK cells with anti-human TLR-neutralizing antibodies versus suitable isotype antibody controls (InvivoGen). Antibody preparations were added to achieve a final concentration of 10 μg/ml. Wells were then supplemented with HEK-Blue detection medium containing the HEK cells and incubated for 16 to 26 h. The plates were read at 630 nm for secretory alkaline phosphatase (SEAP) activity.
Statistical analysis.Statistical tests included calculation of sample means and standard deviations. Histogram bars depict the sample means, and the error bars represent standard deviations (the number of samples analyzed are listed in the figure legends). One-factor analysis of variance (ANOVA) with pairwise comparisons using Dunnett’s multiple-comparison test was then used to determine significant differences between treatment groups and controls.
ACKNOWLEDGMENT
F.C.N. acknowledges grant support from NIH grant 1R01DE027642.
FOOTNOTES
- Received 20 November 2019.
- Returned for modification 12 December 2019.
- Accepted 3 January 2020.
- Accepted manuscript posted online 13 January 2020.
- Copyright © 2020 American Society for Microbiology.
