Inositol-phosphodihydroceramides in the periodontal pathogen Tannerella forsythia: Structural analysis and incorporation of exogenous myo-inositol

Abstract

Background

Unique phosphodihydroceramides containing phosphoethanolamine and glycerol have been previously described in Porphyromonas gingivalis. Importantly, they were shown to possess pro-inflammatory properties. Other common human bacteria were screened for the presence of these lipids, and they were found, amongst others, in the oral pathogen Tannerella forsythia. To date, no detailed study into the lipids of this organism has been performed.

Methods

Lipids were extracted, separated and purified by HPTLC, and analyzed using GC-MS, ESI–MS and NMR. Of special interest was how T. forsythia acquires the metabolic precursors for the lipids studied here. This was assayed by radioactive and stable isotope incorporation using carbon-14 and deuterium labeled myo-inositol, added to the growth medium.

Results

T. forsythia synthesizes two phosphodihydroceramides (Tf GL1, Tf GL2) which are constituted by phospho-myo-inositol linked to either a 17-, 18-, or 19-carbon sphinganine, N-linked to either a branched 17:0(3-OH) or a linear 16:0(3-OH) fatty acid which, in Tf GL2, is, in turn, ester-substituted with a branched 15:0 fatty acid. T. forsythia lacks the enzymatic machinery required for myo-inositol synthesis but was found to internalize inositol from the medium for the synthesis of both Tf GL1 and Tf GL2.

Conclusion

The study describes two novel glycolipids in T. forsythia which could be essential in this organism. Their synthesis could be reliant on an external source of myo-inositol.

General significance

The effects of these unique lipids on the immune system and their role in bacterial virulence could be relevant in the search for new drug targets.

1. Introduction

Tannerella forsythia is a Gram-negative anaerobic oral pathogen, a member of the so-called “red complex” of bacteria that is strongly associated with periodontitis, affecting millions of people worldwide[1-3]. The effects on the periodontium include loss of the alveolar bone around the teeth, swelling and bleeding of the gums and, in more severe cases, loss of teeth. Apart from oral health issues, emerging evidence shows a relationship between periodontal disease, cardiovascular disease and other systemic chronic diseases involving inflammation [4].

Research into bacterial phospholipid synthesis pathways has received attention related to both the study of host-pathogen interactions and the continuously growing interest in new drug targets. In particular, the study of bacterial sphingolipids has recently come into the spotlight as these compounds seem to play an important role in the host-microbe balance and could, therefore, be key to the pathogenesis of diseases [5,6]. In mammalian cells, sphingolipids have been seen to act as signaling molecules and regulators of important processes [7]. Most prokaryotic cells do not contain sphingolipids. However, they have been shown to be present in some bacteria, particularly in anaerobes, including Sphingobacterium, Sphingomonas, Bacteroides, and Bdellovibrio stolpii, some of which are pathogens or parasites able to synthesize their own membrane sphingolipids [8]. Bacterial pathogens which cannot produce sphingolipids, can utilize host sphingolipids to promote their own virulence [5]. A host sphingolipid acquired by a pathogen can be used to evade the immune system and, in turn, microbial sphingolipids incorporated into host membranes could interfere with intercellular signaling and, therefore, impair an immune response to the pathogen or cause an autoimmune response leading to the destruction of host tissues through molecular mimicry [5,9].

Recently, novel phosphorylated dihydroceramide (DHC) lipids were identified in the periodontal pathogen Porphyromonas gingivalis [10]. These include both low-mass and high-mass forms of phosphoglycerol DHC and phosphoethanolamine DHC lipids. The phosphorylated polar head groups are linked to a core lipid structure consisting of either a 17-, 18-, or 19-carbon base in amide linkage to iso-17:0(3-OH) fatty acid (FA). The phosphoglycerol DHC lipids present forms with an additional iso-15:0 FA in ester linkage to the hydroxyl group of the iso-17:0(3-OH) FA. Other bacteria produce lipids that are similar to these but still markedly distinct. Sphingobacterium species synthesize phosphoethanolamine, phosphomannose and phospho-myo-inositol DHCs with the long-chain base in amide linkage to either iso-15:0 or iso-15:0(2-OH) FAs [11]. Neither phosphoglycerol DHCs nor additional ester-linked FAs were reported. In turn, phosphoglycerol ceramide has been found in Bacteroides melaningenicus but with different FA substitutions, containing a variety of saturated and monounsaturated FAs [12], suggesting that Nichols et al. [10] found novel lipids that are distinct from those produced by other organisms. These unusual DHCs have since also been isolated from some Bacteroides, Parabacteroides and Prevotella species of intestinal bacteria and from the oral pathogens T. forsythia and Prevotella intermedia [9]. In addition, they have been found to stimulate pro-inflammatory responses in gingival fibroblasts, enhance autoimmunity, promote apoptosis, and accumulate in diseased gingival tissue and other host tissues distant from the sites normally colonized by the bacteria, hence leading to the hypothesis that they might form the link between periodontal disease and systemic chronic diseases involving inflammation [9,13-15].

In this study, we report novel classes of phosphorylated DHCs which are major cellular lipid components in T. forsythia and have a similar lipid portion to the DHCs described by Nichols et al. [10] but contain phospho-myo-inositol as a polar head-group, instead of phosphoglycerol or phosphoethanolamine. We describe the chemical structure of these glycosphingolipids as elucidated by ESI–MS, GC–MS and NMR, and demonstrate how T. forsythia can internalize radiolabeled myo-inositol from the growth medium and incorporate it into its inositol-containing lipids. The potential relevance of these newly identified unique DHC sphingolipids lies in their structure and presumable function, as inositol-containing sphingolipids, rarely found in bacteria, can serve as powerful signal transduction molecules and, therefore, could have an important role in the pathogen–host interaction [6,16].

2. Materials and methods

2.1. Bacterial strains, medium and culture conditions

T. forsythia strain ATCC 43037 (American Type Culture Collection, Manassas, VA, USA) was grown anaerobically at 37 °C for 4–7 days in either 1-L or 10-mL flasks in brain heart infusion medium (BHI) supplemented with N-acetylmuramic acid, horse serum and gentamicin, as described previously [17].

For radioisotope labeling experiments, T. forsythia was grown for 4 days as described above in a 10-mL batch culture in the presence of 1 μCi of [¹⁴C(U)] myo-inositol (Biotrend, Cologne, Germany), added to the media in 10 μL of ethanol:water (9:1).

For stable isotope labeling experiments, T. forsythia was grown for 4 days as described above in a 250-mL batch culture in the additional presence of 400 μM deuterium labeled myo-inositol-1,2,3,4,5,6-D6 (EQ Laboratories GmbH, Augsburg, Germany) to give a 1:1 ratio between deuterated and endogenous inositol.

2.2. General and analytical methods

All solvents were purchased reagent-grade from Roth or Sigma-Aldrich (both Vienna, Austria), unless stated otherwise.

Lipids were analyzed by HPTLC on nano-silica gel glass plates (Sigma-Aldrich) or HPTLC-silica 60 on aluminum sheets (Merck, Darmstadt, Germany) developed in chloroform:methanol:water (65:25:4), referred to here as solvent system A.

For lipid analysis by mass spectrometry (ESI–MS), lipids were either visualized after HPTLC separation under 365-nm UV-light after staining with a primuline (Sigma-Aldrich) solution [18] and then extracted from the matrix, or purified from the silica plates as described in Section 2.4.

Other lipid visualization techniques used include the multipurpose Hanessian’s stain for total lipid visualization [19] and carbohydrate staining with thymol [20].

Lipids of interest were scratched from the HPTLC plate and the scratched silica matrix was placed in a glass Pasteur pipette packed with glass-wool and the lipids were eluted with chloroform:methanol (1:1). The sample was then filtered through a 0.2-μm PTFE filter (Roth) and dried under a flow of nitrogen.

2.3. Lipid extraction

One gram of lyophilized T. forsythia wild-type biomass was washed extensively with ethanol and acetone prior to lipid extraction in order to remove any lipid contaminants coming from the rich growth medium [21].

Lipid extraction proceeded stepwise following the method established by Bligh and Dyer [22]. Eight mL of water were added to the biomass followed by 30 mL of chloroform:methanol (1:2) and sonicated for 15 min using a probe Branson sonifier (output 4, duty cycle 40%). Subsequently, 10 mL of chloroform were added followed by 10 mL of water, performing sonication for 5 min after each addition. Phase separation was then accelerated using centrifugation (25 min, 10,000 ×g) and the bottom organic phase was separated and filtered through a 0.45-μm PTFE filter followed by a 0.2-μm PTFE filter. The procedure was repeated twice on the solid interphase, and the organic phases were pooled and dried yielding 30 mg of total lipids which were redissolved in chloroform:methanol (8:2) to a concentration of 6 mg/mL.

2.4. Isolation of Tf GL1 and Tf GL2 glycolipids

The glycolipids of interest, named Tf GL1 and Tf GL2, were purified separately from several HPTLC-plates by applying the sample over the full length of the silica plate (50 μg of total lipid extract per cm of baseline). The TLC plates were developed in solvent A, air-dried, and the silica gel was scratched from the glass plate as a band from 1.5 mm below to 1.5 mm above the respective R_f value. The scratched plates were then treated with Hanessian’s stain in order to ensure that Tf GL1 and Tf GL2 had been selectively removed. The lipids were then re-eluted from the combined silica matrix, as described above, filtered and dried under a flow of nitrogen. Lipids were stored at −20 °C until further analysis.

2.5. GC–MS compositional analysis

The composition of Tf GL1 and Tf GL2 was determined after methanolysis (2 M HCl/methanol, 85 °C, 2 h) and acetylation (85 °C, 10 min) by GC–MS. FAs were additionally recovered from the chloroform phase after hydrolysis (4 M HCl, 100 °C, 4 h) and neutralization (5 M NaOH, 100 °C, 30 min) and detected by GC–MS as methyl esters (diazomethane, 10 min, 22 °C). In order to determine the position of the hydroxy groups, after evaporation to dryness, hydroxyl groups of methyl ester-FAs were incubated with BSTFA (4 h, 65 °C) and measured as TMS derivatives. The identification of the detected FAs and carbon chain branching (branched or linear) was achieved based on comparison with external standards of 15:0, 16:0(3-OH) (Larodan AB, Solna, Sweden) and 17:0(3-OH) FAs (Biotrend).

For the determination of the absolute configuration of the hydroxy FAs 16:0(3-OH), 17:0(3-OH) and the most abundant 18:0-sphinganine, Tf GL2 was treated with 48% aqueous HF (4 °C, 48 h) in order to cleave the 1-P-myo-inositol moiety. The remaining ceramide (200 μg) was subjected to methanolysis (2 M HCl/MeOH, 85 °C, 4 h), peracetylated and analyzed by GC–MS by use of d-erythro-sphinganine C18 and l-threo-sphinganine C18 (Sigma-Aldrich) as authentic standards. The obtained sample was further hydrolyzed (4 M HCl, 100 °C, 30 min) and the OH-FAs analyzed by GC–MS after l-phenylethylamide derivatization [23].

GC–MS measurements were performed on an Agilent Technologies 7890A gas chromatograph equipped with a dimethylpolysiloxane column (Agilent Technologies, Santa Clara, CA, USA; HP Ultra 1, 12 m × 0.2 mm × 0.33 μm film thickness) and a 5975C series MSD detector with electron impact ionization (EI) mode under autotune condition at 70 eV. The temperature program was 70 °C for 1.5 min, then 60 °C/min to 110 °C followed by 5 °C/min to 320 °C, held for 10 min.

2.6. ESI–MS analysis

Mass spectrometric measurements of purified lipid samples (see 2.4) were performed on an Esquire 3000^plus 3D-quadrupole ion trap mass spectrometer (Bruker Daltonics, Bremen, Germany) in the negative-ion mode electrospray ionization (ESI–MS). Mass calibration was done with a commercial mixture of perfluorinated trialkyltriazines (ES Tuning Mix, Agilent Technologies, Santa Clara, CA, USA). All analytes were dissolved in methanol hypergrade LC–MS quality (Merck) to a concentration of ~1 mg/mL. Direct infusion experiments were carried out using a Cole Parmer model 74900 syringe pump (Cole Parmer Instruments, Vernon Hills, IL, USA) at a flow rate of 2 μL/min. Full MS-scans were acquired in the range m/z 100–1300 with the target mass set to m/z 1200. Further experimental conditions include: drying gas temperature, 200 °C; capillary voltage, −4 kV; skimmer voltage, 40 V; octapole and lens voltages, according to the target mass set. Helium was used as a buffer gas for full scans. Mass spectra were averaged during a data acquisition time of 0.5 to 1 min and one analytical scan consisted of three successive micro scans resulting in 20 and 40 analytical scans, respectively, for the final mass spectrum. For all low-energy CID-MS² experiments the precursor ion mass was selected at an isolation and activation width of 4 Da. The fragmentation amplitude for dissociating the precursor ions was set to 0.45–0.5 V and the corresponding CID-spectra were collected for at least one minute.

2.7. Nuclear magnetic resonance spectroscopy of Tf GL2

Prior to NMR analyses, purified Tf GL2, 1-P-myo-inositol (Sigma-Aldrich), l-threo-sphinganine C18 (Sigma-Aldrich) and d-erythro-sphinganine C18 (Sigma-Aldrich) were dissolved in deuterated methanol (Deutero GmbH, Kastellaun, Germany) and placed in 5-mm NMR tubes.

All 1D (¹H, ¹³C and ³¹P) and 2D NMR ¹H,¹H COSY, TOCSY as well as ¹H,¹³C HSQC, ¹H,¹³C HMBC (only for Tf GL2), and ¹H,³¹P HMQC experiments performed on Tf GL2 and 1-P-myo-inositol were recorded at 27 °C on a Bruker DRX Avance III 700 MHz spectrometer (operating frequencies of 700.75 MHz for ¹H NMR, 176.2 MHz for ¹³C NMR and 283.7 MHz for ³¹P) and standard Bruker software (TopSpin ver. 3.2). COSY and TOCSY experiments were recorded using data sets (t1 by t2) of 4096 by 512 points, with 6 scans for COSY, and 16 for TOCSY. The TOCSY experiments were carried out in the phase-sensitive mode with mixing times of 120 ms. The ¹H,¹³C correlations measured in the ¹H-detected mode via HSQC-DEPT with proton decoupling in the ¹³C domain and HMBC spectra were acquired using data sets of 4096 by 512 points and 64 scans (HSQC) and 80 scans (HMBC) for each t1 value. HMBC spectra were adjusted to J coupling constant value of 145 Hz and long range proton carbon coupling constant of 10 Hz. Chemical shifts were reported relative to internal methanol (δ_H 3.34, δ_C 49.86) [24] or external phosphoric acid (δ_P 0.0) [25].

2.8. Incorporation of radiolabeled myo-inositol into T. forsythia lipids

The study of radiolabeled inositol incorporation was based on procedures described previously with substantial modification[26-28]. T. forsythia was grown in the presence of 1 μCi of [¹⁴C(U)] myo-inositol (specific radioactivity 200–250 mCi/mmoL) as described in Section 2.1. After 4 days of growth, the biomass was pelleted and lipids were extracted using the Bligh–Dyer method [22]. 80 μL of water were added to the wet biomass followed by 300 μL of chloroform:methanol (1:2), vortexed for 5 min and left shaking for an additional 10 min. Subsequently, 100 μL of chloroform were added followed by 100 μL of water, vortexing for 5 min and left shaking for 10 min, after each addition. After phase separation, the top aqueous phase was carefully removed. The bottom organic phase was then backwashed three times by addition of a 100 μL of water. The radioactivity in each wash fraction and in the organic phase was determined using a liquid scintillation analyzer Tri-Carb 2910 TR (Perkin Elmer, Waltham, MA, USA). The organic phase containing the extracted lipids was dried under a flow of nitrogen, redissolved in chloroform:methanol (8:2), spotted onto a HPTLC silica-gel aluminum-back plate (Merck) and then chromatographed in solvent A. The plate was then subjected to autoradiography using Kodak BioMax XAR film (Sigma-Aldrich) for three weeks at −80 °C.

2.9. Determination of free D₆-labeled and endogenous inositol in the culture medium

The determination of free deuterated and natural inositol in the medium was performed by RP–PGC–TOFMS, using known amounts of pure myo-inositol (Sigma-Aldrich) as standards. The medium samples were prepared by centrifugal filtration (60 min, 10 000 rpm, 4 °C; Amicon Ultra, 3 kDa MW cut-off) and diluted 1:100 in LC–MS grade H₂O (Sigma Aldrich) prior to injection. A detailed description of the chromatographic separation is provided elsewhere [29]. The chromatography column was coupled on-line to a time-of-flight mass spectrometer (Agilent 6230 LC–TOFMS) via a Dual AJS ESI spray chamber. ESI was performed in negative mode (45 psig nebulizer pressure, 350 °C sheath gas temperature, 11 L/min sheath gas flow, 250 °C drying gas temperature, 12 L/min drying gas flow, 110 V fragmentor voltage, 60 V skimmer voltage, 3500 V capillary voltage). MS data was acquired in the mass range 50-1000 m/z at an acquisition rate of 3 spectra per second. Both unlabeled and D₆-labeled myo-inositol were detected as formic acid adducts ([M + COO⁻]⁻, 225.0616 m/z and 231.0993 m/z, respectively) at a retention time of 3.9 min. After peak integration (Agilent MassHunter Quantitative Analysis 7.0, extraction window ±5 ppm), the peak areas were used to calculate the ratio between D₆-labeled and unlabeled inositol.

3. Results

3.1. Isolation of Tf GL1 and TF GL2

HPTLC of the Bligh–Dyer extracts from T. forsythia cells followed by a first screening of the primuline-stained bands using MALDI–MS (data not shown) allowed for the identification of the most abundant lipids present (phosphatidylethanolamine, monogalactosyl diacylglycerol, phosphatidic acid and cardiolipin) as well as of the phosphorylated DHC lipids previously reported by Nichols et al. [9]. In addition, we detected two novel DHC lipids which could be resolved with the solvent system A. The less abundant, more polar lipid, designated as Tf GL1, migrated with an R_f value of 0.10 and the second, more abundant lipid, designated here as Tf GL2, migrated slightly in front of Tf GL1 with an R_f value of 0.15 (Fig. 1A).

Fig. 1.

HPTLC of T. forsythia lipids developed with the solvent system chloroform: methanol:water (65:25:4). Visualization was performed with the Hanessian’s stain for lipids (blue) and with thymol for carbohydrates (pink). Abbreviations: B., baseline; S. f., solvent front. A. Total lipid extract obtained using the Bligh–Dyer extraction method. 30 μg of sample were spotted at the baseline. TF GL1 and Tf GL2 ran at R_f values of 0.10 and 0.15, respectively. B. Purified Tf GL1 and Tf GL2 appeared as single bands in HPTLC and stained positive for carbohydrates. Lane 1: 3 μg of sample, lane 2: 10 μg of sample.

The Tf GL1 and Tf GL2 lipids were subsequently purified from several HPTLC-plates by applying the sample over the full length of the silica plate and scratching the silica from the glass plates at the corresponding R_f values. Fig. 1B shows purified Tf GL1 and Tf GL2 fractions after elution from the silica gel. Both T. forsythia lipids stained positive for carbohydrates as assayed with thymol. Two mg of purified Tf GL2 and 400 μg of purified Tf GL1 were recovered from the total lipid extract, representing ~20% and ~5% of the total lipid fraction, respectively.

3.2. GC–MS compositional analysis of purified Tf GL1 and Tf GL2 glycolipids

Compositional analyses of Tf GL1 and Tf GL2 revealed the presence of inositol, linear 16:0(3-OH) FA, branched 17:0(3-OH) FA as well as sphinganine with a long-chain-length of 17, 18 and 19 carbon atoms. Tf GL2 contained additionally a branched 15:0 FA. The most prominent signals belonged to 17:0(3-OH) FA and 18:0 sphinganine. GC–MS spectra of the sample after methanolysis also contained low intensity signals for glycerol, 16:0, 18:1 and 18:0 FA. However, further analyses by ESI–MS and NMR (see Sections 3.3. and 3.4. below) did not provide evidence that the aforementioned compounds were a structural part of Tf GL1 and Tf GL2. The GC–MS analyses performed on the dephosphorylated Tf GL2 revealed that the 18:0 sphinganine possessed an erythro configuration and both hydroxy FAs were R configured.

3.3. ESI–MS analyses of purified Tf GL1 and TF GL2 glycolipids

Negative-ion ESI–MS spectra of purified Tf GL1 and Tf GL2 (Fig. 2) revealed five [M–H]⁻ ions for each lipid sample corresponding to the deprotonated lipid species of M₁–M₅, where M stands for the neutral mass of a given molecular species. Tf GL1 consisted mainly of the deprotonated ions of M₁, m/z 768; M₂, m/z 782; M₃, m/z 796; M₄, m/z 810; and M₅, m/z 824. The corresponding signals for Tf GL2 were M₁, m/z 992; M₂, m/z 1006; M₃, m/z 1020; M₄, m/z 1034; and M₅, m/z 1048. The spacing of 14 Da between the M₁–M₅ species is indicative of variations in FA chain length by single [−CH₂−] groups, which is typical for bacteria.

Fig. 2.

High-mass region of the negative-ion ESI-mass spectra ([M–H]⁻, where M stands for the neutral mass of a given molecular species) of Tf GL1 (A) and Tf GL2 (B).

Other ions present were the deprotonated sodium chloride adducts of M₃ and M₄. Both samples revealed a markedly similar quantitative distribution of M₁–M₅, with M₄ being the most predominant. The mass difference between the ions of Tf GL1 and Tf GL2 was 224 Da, which could correspond to a missing 15:0 FA in the Tf GL1 structure as compared to Tf GL2, indicating that Tf GL1 and Tf GL2 could be low- and high-mass forms of a single lipid class.

3.3.1. CID–MS² of Tf GL1 and Tf GL2

CID–MS² of Tf GL1 revealed major product ions for all four of the [M_2–5–H]⁻ precursor ions (Fig. 3, Table 1). The resulting spectrum for each parent ion included peaks corresponding to the loss of inositol ± H₂O, [M–H – (162 / 180)]⁻, loss of an amide-linked 16:0(3-OH) or 17:0(3-OH) FA, [M–H – (254 or 268)]⁻, and loss of the inositol and partial loss of the latter FA forming a McLafferty ion, [M–H – (212 or 226)]⁻ [30].

Fig. 3.

Negative-ion ESI low-energy CID–MS² of the Tf GL1 [M–H]⁻ precursor ions: m/z 782 (A), m/z 796 (B), m/z 810 (C) and m/z 824 (D).

Table 1.

Combined GC–MS and ESI–MS data. Major losses and product ions of the [M–H]⁻ precursor ions of the sphingo-phosphoinositols of Tf GL1 (A) and the corresponding O-acyl-sphingo-phosphoinositols of Tf GL2 (B). An asterisk (*) indicates those sphingolipid species containing additionally two diagnostic low-mass product ions for phosphoinositol at m/z 259 and at m/z 241 which are not detected for the O-acyl-sphingo-phosphoinositols (Tf GL2) due to the low mass cut-off of the 3D-ion trap instrument utilized for the experiments. Abbreviations: FA, fatty acid; Sph, sphinganine.

Sphingolipid species	[M–H]−	−162/−180 (loss of inositol)	−242 (loss of the O-linked 15:0 FA)	−(242 + 162/+180) (loss of inositol plus the O-linked 15:0 FA)	Partial loss of the N-linked OH–FA (McLafferty rearrangement) including additional loss of inositol	Loss of the N-linked OH–FA
A
*C33;Sph: 17:0; (N)HO–FA: 16:0 linear	782	620/602	–	–	570 (−180:390)	528
*C34;Sph: 17:0; (N)HO–FA: 17:0 branched	796	634/616	–	–	570 (−180:390)	528
*C34;Sph: 18:0; (N)HO–FA: 16:0 linear	796	634/616	–	–	584 (−180:404)	542
*C35;Sph: 18:0; (N)HO–FA: 17:0 branched	810	648/630	–	–	584 (−180:404)	542
*C36;Sph: 19:0; (N)HO–FA: 17:0 branched	824	662/644	–	–	598 (−180:418)	556
B
C33;Sph: 17:0; (N)HO–FA: 16:0 linear, (O)FA: 15:0 branched	1006	844/826	764	602/584	–	528
C34;Sph: 17:0; (N)HO–FA: 17:0 branched, (O)FA: 15:0 branched	1020	858/840	778	616/598	–	528
C35;Sph: 18:0; (N)HO–FA: 17:0 branched, (O)FA: 15:0 branched	1034	872/854	792	630/612	–	542
C36;Sph: 19:0; (N)HO–FA: 17:0 branched, (O)FA: 15:0 branched	1048	886/868	806	644/626	–	556

CID–MS² of [M₃–H]⁻ for Tf GL1 proved the parent ion to be a mixture of two separate species of equal mass as both m/z 570 ([M–H – 226]⁻) and m/z 584 ([M–H – 212]⁻) McLafferty ions were present amongst the product ions indicating the presence of both amide-linked 16:0(3-OH) and amide-linked 17:0(3-OH) FAs. In addition, complete loss of this amide-linked hydroxy-FA revealed two separate ions (m/z 528 and 542) indicating that carbon chain length variations occurred also in the remaining core-lipid structure.

CID–MS² of Tf GL2 also revealed product ions for all four of the [M_2–5–H]⁻ precursor ions (Fig. 4, Table 1). The detected product ions corresponded to the loss of inositol ± H₂O, [M–H – (162 / 180)]⁻, but did not result in the previous McLafferty ions but in a loss of an ester-bonded 15:0 FA, forming a double bond in the remaining structure, [M–H – 242]⁻, partially accompanied by the additional loss of the inositol ± H₂O, [M–H – (404 / 422)]⁻. Loss of the ester-bonded 15:0 FA was followed by a full loss of the amide linked 16:0(3-OH) or 17:0(3-OH) FA [M–H – (478 or 492)]⁻, which are 3-hydroxy FAs as determined by GC–MS and the McLafferty ions formed in the CID–MS² of Tf GL1.

Fig. 4.

Negative-ion ESI low-energy CID-MS² of the Tf GL2 [M–H]⁻ precursor ions: m/z 1006 (A), m/z 1020 (B), m/z 1034 (C) and m/z 1048 (D).

The lower molecular-mass species, Tf GL1, clearly showed two diagnostic low mass product ions for phospho-inositol at m/z 241 and 259 (which could not be found in Tf GL2 due to the low mass cut-off of the 3D-ion trap MS utilized) and presented peaks at m/z 528, 542 and 556 which also appeared as minor product ions in the higher molecular-mass species. These ions could be assigned to a sphinganine (dihydroceramide)-phospho-inositol structure where the sphinganine presents variations in the chain length assigned as 17:0 sphinganine, 18:0 sphinganine and 19:0 sphinganine, respectively.

All resulting major precursor and product ions are summarized in Table 1, along with how the variations in carbon chain lengths of both the sphinganine and the amide-linked 3-hydroxy FAs result in the lipid species M₂ to M₅. The most abundant M₄ for both Tf GL1 and Tf GL2 corresponded to the species having 18:0 sphinganine and the amide-linked 17:0(3-OH) FA, in good agreement with the GC–MS data. Fig. 5 shows the structure of Tf GL1 and Tf GL2, as elucidated from the combined GC–MS and ESI–MS data, and the described fragmentation pathways in CID–MS².

Fig. 5.

Structure (A) and general fragmentation pathways of [M–H]⁻ precursor ions of O-acyl-sphingo-phosphoinositols (Tf GL2) and O-deacyl-sphingo-phosphoinositols (Tf GL1) (B).

3.4. NMR analyses of Tf GL2

The chemical nature of the more abundant Tf GL2, composed of phospho (P)-inositol linked to erythro-sphinganine, further substituted by amide-linked hydroxy FAs and an ester-linked 15:0 FA, could be confirmed by NMR analyses (Fig. 6, Table 2). The P-inositol was identified as 1-P-myo-inositol based on the chemical shifts of the authentic standard (atoms 1–6; Fig. 6). The ¹H spectrum of Tf GL2 showed, in addition to the signals of 1-P-myo-inositol, the downfield shifted multiplet at 5.27 ppm (due to the strong deshielding effect), signals from the α-methylene groups of the amide- and ester-bound FAs (2.54, 2.35 ppm), signals resonating in the region from 1.68 to 1.61 ppm (γ-methylene group of the amide bound FA and methylene group at position 4 of the sphinganine base), and the groups of signals from 1.41 to 1.27 ppm and from 0.95 to 0.88 ppm corresponding to −CH₂− and terminal methyl groups of the alkyl chains, respectively. Scalar correlation of δ_H 5.27 to δ_C 172.98 and 175.78 in the HMBC spectrum enabled assignment of the signal to the β-hydroxy group of the amide linked FA, being further O-acylated. The ³¹P signal at 1.84 ppm correlated with the signals at 3.92 ppm (position 1 of inositol) and 4.26 and 3.97 ppm (position 1 of the sphinganine base) as seen in the ¹H,³¹P HMQC experiment, identifying the position of the phospho-ester linkage. A HSQC-DEPT experiment (Fig. 6, Table 2) provided the direct correlation of all assigned ¹H with ¹³C signals. Comparing ¹H spectra of the Tf GL2 with the authentic standards of the erythro and threo forms of sphinganine (that differ in the J_2,3 coupling constants, being ~ 6 Hz for erythro and ~2 Hz for threo, respectively [31]), enabled the identification of an erythro-sphinganine in Tf GL2. Taken together, NMR analysis of Tf GL2 confirmed the structure to be composed of 1-P-myo-inositol phosphoester-linked to an erythro-sphinganine base which is substituted via the amide group to an R-hydroxy FA, in turn estersubstituted by a second FA.

Fig. 6.

Overlay of the ¹H NMR spectra of 1-P-myo-inositol (black) and Tf GL2 (grey), along with the overlay of the 2D experiments HSQC-DEPT (grey dots) and HMBC (black dots). Numbers indicate the position in the structure of Tf GL2 (top, left). Erythro-sphinganine is drawn here in its d (8S, 9R) absolute configuration, however d- or l-configuration could not be distinguished.

Table 2.

¹H, ¹³C and ³¹P NMR assignments of 1-P-myo-inositol and Tf GL2 and scalar correlations seen in the HMBC spectrum. Numbers in bold indicate the position in the structure of Tf GL2 as shown in Fig. 6. Multiplicity abbreviations: a, b, hydrogen atoms; d, doublet; t, triplet; m, multiplet; dd, double doublet.

	1H (ppm)	13C (ppm)	31P (ppm)	HMBC (ppm)
1-P-myo-Inositol
1	m 3.95	78.21	2.22
2	t 4.22	73.95	–
3	m 3.42	73.68	–
4	t 3.66	74.72	–
5	t 3.23	77.11	–
6	t 3.82	74.28	–
Tf GL2
1	m 3.92	79.33	1.84
2	t 4.23	73.93	–
3	dd 3.41	73.68	–
4	t 3.65	74.83	–
5	t 3.22	77.03	–
6	t 3.78	74.05	–
7a	m 4.26	66.69	1.84
7b	m 3.97
8	m 3.86	56.43	–	8 3.86/11 172.98
9	m 3.68	71.45	–
10a, b	1.68–1.61	27.02	–
11	–	172.98	–	172.98
12	d 2.54	42.94	–	12 2.54/13 73.19
				12 2.54/11 172.98
13	m 5.27	73.19	–	13 5.27/11 172.98
				13.5.27/15 175.78
14a, b	1.68–1.64	27.02	–	14 1.66/15 175.78 (faint signal)
15	–	175.78	–	175.78
16a, b	t 2.35	36.12	–	16 2.35/15 175.78 (strong signal)

3.5. Uptake of the glycolipid metabolite myo-inositol from the medium

A BLASTP search against the proteome of T. forsythia ATCC 43037 did not identify proteins homologous to the two enzymes involved in inositol synthesis, i.e. inositol-3-phosphate synthase and inositolmonophosphatase. This suggests that T. forsythia utilizes exogenous myo-inositol for the synthesis of Tf GL1 and Tf GL2. In order to prove the uptake of the glycolipid precursor myo-inositol from the medium, T. forsythia was grown in the presence of [¹⁴C(U)] myo-inositol. Lipids were extracted using the Bligh–Dyer method and the organic phase was backwashed several times with water to insure complete removal of non-incorporated radiolabeled inositol from the organic phase. The latter fraction (containing about 0.2% of the radioactive precursor initially added) was spotted onto a HPTLC plate and separated using solvent system A. After a 3-week exposure time, two bands were clearly visible at the exact R_f values of 0.10 and 0.15 corresponding to Tf GL1 and Tf GL2, respectively (Fig. 7). This indicated that T. forsythia is able to import inositol from the medium and incorporate it into its lipids. No other bands were detected, indicating that [¹⁴C(U)] myo-inositol was solely incorporated into Tf GL1 and Tf GL2.

Fig. 7.

Autoradiography after HPTLC analysis of a total lipid extract of T. forsythia grown in the presence of [¹⁴C(U)] myo-inositol. Lane 1 shows free [¹⁴C(U)] myo-inositol which does not migrate in the solvent system used. Lane 2 shows the total lipid extract containing two lipid bands with incorporated [¹⁴C(U)] myo-inositol with R_f values of 0.10 and 0.15 corresponding to Tf GL1 and Tf GL2, respectively. Abbreviations: B., baseline; S. f., solvent front; R_f, retardation factor.

3.6. Tf GL2 labeling with D₆-labeled myo-inositol

Quantification of free inositol in the medium revealed its presence at an approximate concentration of 400 μM, as estimated by using an external myo-inositol standard. D₆-labeled myo-inositol was then added to the medium before inoculation to give a ratio between deuterated and unlabeled inositol of close to 1:1. The ratio at the beginning of the culture and after harvesting the biomass was experimentally determined as 1.05 and 1.03, respectively. Tf GL2 was then purified and subjected to ESI–MS analysis. The ratio between deuterated and unlabeled Tf GL2 was found to be 1.11 (Fig. 8), demonstrating that the isotopic distribution of D₆-labeled inositol in Tf GL2 was practically equal to that in the medium. This provides evidence that the incorporated inositol is exclusively derived from the medium.

Fig. 8.

High-mass region of the negative-ion ESI-mass spectra of Tf GL2 isolated from T. forsythia grown in medium containing deuterium labeled and unlabeled inositol at a ratio of close to 1:1. Peaks labeled with [d₆M–H]⁻ correspond to ions derived from deuterated inositol labeled Tf GL2, as indicated from the mass difference of 6 Da from the corresponding Tf GL2 [M–H]⁻ ions.

4. Discussion

As a close relative of Bacteroides, the ability of T. forsythia to synthesize sphingolipids, an attribute rarely found amongst bacteria, is phylogenetically well justified [6]. Although the lipidome of this bacterium is poorly characterized, some of its sphingolipids are unique phosphorylated DHCs, prevalent amongst a set of common human bacteria and biologically relevant due to their putative role in eliciting an immune response [9,10,13].

In this study, we report on two new phosphorylated DHCs (Tf GL1 and Tf GL2) which are major cellular lipid components of T. forsythia, structurally similar to those already described [10], but containing myo-inositol as a polar head-group. The structure of these unusual glycolipids was elucidated by a combination of ESI–MS, GC–MS and NMR. Tf GL1 and Tf GL2 consist of phospho-myo-inositol linked to a core lipid structure consisting of either a 17-, 18-, or 19-carbon sphinganine base in amide linkage to either branched 17:0(3-OH) or linear 16:0(3-OH) FA which, in the higher molecular weight structure of Tf GL2, is, in turn, substituted with an additional branched 15:0 FA in ester linkage. The lipid core described here is similar to that described by Nichols et al. [10]. These authors reported only 17:0(3-OH) FA to be present in amide linkage to sphinganine. In addition, the 17:0(3-OH) FA, 15:0 FA and 17- and 19-carbon sphinganine were determined to be iso-branched (the 18-carbon sphinganine was reported to be linear), as determined by comparison with the synthetic standards.

It has been postulated that there exists a relationship between periodontal disease and other systemic chronic diseases where inflammation seems to be the common link [4]. The novel DHCs published by Nichols et al. [10] were seen to be present in host tissues distant from the sites of infection and evoke an inflammatory response in vitro [9]. Their FA substitutions are thought responsible for this observation as the lipid portion is what differs structurally from other DHCs. The finding that these core lipid structures also exist with a rare and important polar head group like inositol, often involved in protein–lipid interactions on cell surfaces [32], further increases the interaction possibilities and widens the spectra of effects this type of lipid could have on the immune system, especially when considering that inositol-containing lipids and sphingolipids, although rare in microorganisms, seem more prevalent amongst pathogenic bacteria [5,6,33]. Confirmed functional information on inositol-containing structures is known for just a few organisms [32]. The use of myo-inositol for the synthesis of phosphatidylinositol (PI) and other lipids is essential in eukaryotes, including pathogenic fungi and protozoa, but is rare amongst bacteria and has been described only for a small group of actinobacteria (including mycobacteria, Streptomyces and corynebacteria), Bacteroidetes (Sphingobacterium) and in myxobacteria and Treponema [33,34]. Further occurrence of inositol in bacteria includes the use of this metabolite in Mycobacteria for the synthesis of mycothiol, or in other bacteria to produce di-myo-inositol-phosphate, or it can be used as a precursor for various aminoglycoside/aminocyclitol antibiotics [34-37]. A number of bacteria can internalize myo-inositol and use it as a nutrient source but cannot incorporate it into biosynthetic pathways [33,38,39].

Not only is it unique to find T. forsythia with inositol containing lipids but it is of particular interest how the pathogen acquires this basic molecule for the synthesis of those. For the synthesis of PI, organisms either synthesize myo-inositol from glucose-6-phosphate or import it from the host [33]. There is a universal mechanism for the synthesis of inositol which requires the combined action of two enzymes, inositol-3-phosphate synthase (named Ino1 in yeast) and inositolmonophosphatase, both of which are highly conserved proteins between the different domains of life [27,33]. Genes homologous to ino1 have been found in related Bacteroides sp. and in the uncultured, health-associated Tannerella sp. BU063 [40], although the physiological significance of inositol synthesis in these bacteria is not clear [34]. In contrast, no genes predicted to be involved in inositol synthesis could be identified in the genome of T. forsythia [41]. By adding radiolabeled myo-inositol to the medium of T. forsythia during growth, we were able to demonstrate that the organism is able to internalize this compound and incorporate it into its lipids. Moreover, labeling experiments with deuterated inositol indicated that the source of this metabolite is solely exogenous.

In some bacteria, such as in mycobacteria, PI is further built up with other sugar residues, such as mannose, to form phosphatidyl-myo-inositol mannosides (PIMs) [42]. In our study, no other lipids were seen to incorporate labeled myo-inositol, and so further derivatization of the inositol-containing lipids to form other major lipid components is not expected to occur in T. forsythia.

Like other bacteria residing in human hosts, T. forsythia has adapted to better suit its niche and has optimized its physiology to better use the nutrients and resources available, minimizing the need of de novo synthesis of vital components. Furthermore, this ability in T. forsythia may be a key to pathogenicity. For example, when comparing this virulent strain to the health-associated Tannerella sp. BU063, the latter presents two de novo synthesizing enzymes required to convert UDP-N-acetyl-d-glucosamine to UDP-N-acetylmuramic acid, enzymes which lack in T. forsythia [40]. N-Acetylmuramic acid can be presumably scavenged by T. forsythia from the host in its natural habitat but has to be added to the media for growth under laboratory conditions [43]. Similarly, T. forsythia, unlike the harmless Tannerella sp. BU063, does not have any gene homologs to ino1, involved in inositol synthesis, but can import inositol from the medium for synthesis of major lipid components. Human serum, contained for example in the gingival crevice, contains 61 ± 12 μM of myo-inositol which the pathogen could exploit for this purpose [33] and hence, taken together, we hypothesize that the uptake of inositol could be a contributing factor to virulence.

In mycobacteria, internalized inositol is first phosphorylated in order to be used as a substrate for the formation of phosphatidylinositol phosphate (PIP), formed as an intermediate during the formation of PI. This led to a newly proposed mechanism for synthesis of inositol-containing lipids in bacteria, different to that in eukaryotes [27]. Here, exogenous inositol is imported via an inositol or a sugar transporter and then phosphorylated to form myo-inositol-1-phosphate which can be coupled to CDP-diacylglycerol to form the intermediate PIP, by a CDP-alcohol phosphatidyltransferase. An added dephosphorylation step would then be needed to convert PIP to PI. As in mycobacteria[16,35], no obvious transporter genes or candidates for an inositol kinase could be identified in T. forsythia. Jorge et al. also described a new biosynthetic pathway restricted to bacteria which does not rely on CTP-activated glycerolipids but uses instead activated inositol in the form of CDP-inositol [44]. This pathway was elucidated for Rhodothermus marinus (phylum Bacteroidetes) and implies the inositol-3-phosphate synthase gene to be located within an operon-like structure together with a l-myo-inositol-1-phosphate cytidylyltransferase (IPCT) and a bacterial dialkylether glycerophosphoinositol synthase (BEPIS). This operon is found in a number of different bacterial genera and a BEPIS homologue can also be found in T. forsythia (Tanf_07760), but the other members of the operon are not present; especially the IPCT would be needed for synthesis of activated CDP-inositol from the imported metabolite. Therefore, in this context, the physiological importance of this putative BEPIS is currently not clear.

In order to import inositol from the medium in sufficient quantity to support the synthesis of such major lipid components, representing ~25% of its membrane lipids, we would expect a designated inositol transporter to be present. The major transporter of myo-inositol in Bacillus subtilis strains and in Lactobacillus casei BL23 was found to be encoded by the iolT gene which forms part of an operon for inositol catabolism, as inositol in these organisms can be used as a nutrient source [38,39]. As done for Mycobacterium tuberculosis [35], iolT sequences were used in a BLASTP search against the proteome of T. forsythia ATCC 43037. Only one significant hit could be found, Tanf_07060, classified as a major facilitator superfamily (MFS) transporter of the sugar porter (SP) family. However, this gene lies within a possible operon for the uptake and use of xylose. Therefore, the capacity of this protein to act as an inositol transporter has to be verified experimentally which is now subject of further studies. Identification of the responsible gene and creation of the corresponding knock-out mutant might give insight into how essential the utilization of inositol is in this pathogen.

5. Conclusion

The study performed here on the lipids of T. forsythia, describes the novel structures of two classes of phospho-myo-inositol-DHCs. Considering that these lipids make up such a large portion of the membrane, they are possibly essential in this oral pathogen which most likely obtains all of the myo-inositol needed from the surrounding media. The effects of these lipids on the host immune system and its implication in the virulence of this pathogen are yet to be evaluated but could help get functional information on inositol-containing lipids and could be relevant in the search for new drug targets.

Abbreviations

BSTFA

N,O-bis(trimethylsilyl)trifluoroacetamide

CID

collision-induced dissociation

COSY

correlation spectroscopy

DHC

dihydroceramide

ESI–MS

electrospray ionization mass spectrometry

FA

fatty acid

GC–MS

gas chromatography coupled with mass spectrometry

HMBC

heteronuclear multiple-bond correlation spectroscopy

HPTLC

high-performance thin layer chromatography

HSQC–DEPT

heteronuclear single quantum coherence–distortionless enhancement by polarization transfer

MALDI–MS

matrix-assisted laser desorption/ionization mass spectrometry

NMR

nuclear magnetic resonance spectroscopy

PDHC

phosphorylated dihydroceramide

PI

phosphatidylinositol

PIP

phosphatidylinositol phosphate

PTFE

polytetrafluoroethylene

TMS

tetramethylsilane

TOCSY

total correlation spectroscopy