Novel Glycerophospholipid, Lipo- and N-acyl Amino Acids from Bacteroidetes: Isolation, Structure Elucidation and Bioactivity

The ‘core’ metabolome of the Bacteroidetes genus Chitinophaga was recently discovered to consist of only seven metabolites. A structural relationship in terms of shared lipid moieties among four of them was postulated. Here, structure elucidation and characterization via ultra-high resolution mass spectrometry (UHR-MS) and nuclear magnetic resonance (NMR) spectroscopy of those four lipids (two lipoamino acids (LAAs), two lysophosphatidylethanolamines (LPEs)), as well as several other undescribed LAAs and N-acyl amino acids (NAAAs), identified during isolation were carried out. The LAAs represent closely related analogs of the literature-known LAAs, such as the glycine-serine dipeptide lipids 430 (2) and 654. Most of the here characterized LAAs (1, 5–11) are members of a so far undescribed glycine-serine-ornithine tripeptide lipid family. Moreover, this study reports three novel NAAAs (N-(5-methyl)hexanoyl tyrosine (14) and N-(7-methyl)octanoyl tyrosine (15) or phenylalanine (16)) from Olivibacter sp. FHG000416, another Bacteroidetes strain initially selected as best in-house producer for isolation of lipid 430. Antimicrobial profiling revealed most isolated LAAs (1–3) and the two LPE ‘core’ metabolites (12, 13) active against the Gram-negative pathogen M. catarrhalis ATCC 25238 and the Gram-positive bacterium M. luteus DSM 20030. For LAA 1, additional growth inhibition activity against B. subtilis DSM 10 was observed.


Introduction
Lipids are a diverse group of natural biomolecules. Thousands of distinct lipids, such as glycerolipids, sterol lipids, sphingolipids, lipoamino acids (LAAs), and phospholipids, are ubiquitous in all organisms. Each of them is chemically unique, and they exhibit different biological functions. Given the diversity in both the chemical and physical properties of lipids and the fact that each lipid type is involved at various stages of cellular processes, the definition of lipid function besides their primary biological role, i.e., the formation of cell membrane matrixes, is challenging. Described functions in cellular signaling, energy storage, or an implication as substrate for metabolite or protein lipidation are only examples [1].
For enabling doubtless structure elucidation, various analytical methods such as MS and NMR are well established and intensively used in the field of natural product research in general. For lipids as one important sub-class, first, gas or liquid chromatography coupled to diverse mass spectrometry methods represent two key analytical techniques [37][38][39]. Second, NMR spectroscopy is not only widely utilized for structure elucidations of single lipids, successful applications regarding qualitative and quantitative analysis of lipids in complex mixtures are also reported in this context [40][41][42][43][44][45].
In a previous study, we identified the 'core' metabolome of the Bacteroidetes genus Chitinophaga, consisting of only seven metabolites [46]. We postulated a structural relationship among four of them based on their MS/MS fragmentation pattern, which suggested them to be unknown LAAs and LPEs. Here, we report the identification in total of 16 diverse lipids (11 LAAs, 2 LPEs, 3 NAAAs) produced by Chitinophaga spp. and Olivibacter sp. FHG000416, the 'core' ones included. Isolation and structure elucidation of nine of those lipids isolated from Chitinophaga eiseniae DSM 22224 [47] and FHG000416 was successfully achieved. Based on the four Chitinophaga core lipids, the known lipid 430, several novel LAA analogs thereof, and three undescribed NAAAs were characterized. In this context, we identified a novel glycine-serine-ornithine LAA family that is closely related to the previously described glycine and glycine-serine LAA families that lipid 430 and 654 belong to. Investigating them for their antimicrobial activity revealed a common growth inhibition effect against the Gram-negative pathogen M. catarrhalis ATCC 25238 and the Gram-positive bacterium M. luteus DSM 20030, with LAA 1 as the most potent one, also showing growth-inhibiting activity against B. subtilis DSM 10.

Lipoamino Acids
A previous study revealed a small core metabolome of the Bacteroidetes genus Chitinophaga, identified as a genus of underexplored talented producers of natural products.  (Table S1). Based on their MS/MS fragmentation pattern, similar lipid moieties were postulated, suggesting a structural relationship [46].
For the first core metabolite, the molecular formula C 27 H 52 N 4 O 7 (1) was assigned according to the UHR-ESI-MS ion peak at m/z 545.3908 [M + H] + . The MS/MS fragmentation of 1 revealed neutral losses of H 2 O, followed by the three amino acids ornithine, serine, and glycine, resulting in the fragment ion of m/z 251.2366 [M + H] + . The corresponding ion formula of C 17 H 31 O + indicated a fatty acyl group based on the molecular composition and apparent carbon-to-hydrogen ratio ( Figure 1A). A missing hit in a database search indicated the potential novelty of LAA 1, containing a glycine-serine-ornithine tripeptide. However, very close relatives, such as lipid 430 (2) and lipid 654, belong to the glycine-serine dipeptide LAA family, known to be biosynthesized by several bacteria, including representatives of the phylum Bacteroidetes [14,30,34]. Therefore, the metabolomics data generated in our previous study was examined for the presence of both dipeptide lipids. A corresponding ion peak for lipid 430 (2) at m/z 431.3114 [M + H] + with a RT of 13.0 min was identified in 23 of the analyzed 25 Chitinophaga metabolomes (Table S1). Similar serial neutral losses of H 2 O, serine, and glycine, together with the same remaining fragment ion of m/z 251.2367 [M + H] + , strongly suggested 1 being a close derivative of 2, expanded by an additional ornithine residue ( Figure 1B). To provide a sufficient amount of the compounds for structure confirmation via NMR, our in-house Bacteroidetes-based extract library was examined for enhanced production of 1 and 2. Lipid 430 (2) was enriched in extracts of strain FHG000416, which was assigned by 16S rRNA sequence analysis to the Bacteroidetes genus Olivibacter due to a sequence identity of~94.5% towards Olivibacter domesticus DSM 18733 [48,49]. Extracts from both strains, C. eiseniae DSM 22224 and FHG000416, were considered as starting points for compound isolation. NMR analysis confirmed compound 2 to be lipid 430 since the data were in good agreement with the literature (Table 1) [35]. Compared thereto, 1 H and 2D spectra of 1 were highly similar. Accordingly, the fatty acyl motif linked to glycine and serine was identified as iso-heptadecanoic acid (iso-C 17:0 ). An additional amide proton 2 -NH (δ H 7.66-7.59 ppm) was observed showing COSY correlation to methine proton H-2 (δ H 3.82-3.75 ppm, δ C 53.6 ppm). Further correlations to methylene protons H-3 , H-4 , and H-5 confirmed the presence of a C-terminal ornithine residue, as postulated based on the MS/MS fragmentation, identifying lipid 1 as 5-amino-2-(3-hydroxy-2-(2-(3hydroxy-15-methylhexadecanamido)acetamido)propanamido)pentanoic acid ( Figure 2; Table 1).

Phospholipids
In addition to the aforementioned LAAs, dereplication of the last two Chitinophaga core buckets with m/z 452. 2769  , indicating the structural variance to be located in the acyl motif. For NMR analysis, the isolation of these compounds was again carried out from extracts of Olivibacter sp. FHG000416, as higher production titers were observed compared with C. eiseniae DSM 22224. For 12, NMR data confirmed the occurrence of lysophosphatidylethanolamine (C 16:1 ), as postulated by dereplication. According to published data for LPE 451, the double bond (δ H-15/16 5.24-5.12 ppm) of the mono-unsaturated palmitoyl motif was assigned to ∆ 9 position with Z configuration, with the acyl chain being attached to the glycerol moiety at sn-1 position (Table S3) [51]. Thus, 12 was identified as 1-(9Z-palmitoyl)-2-hydroxy-sn-glycerol-3-phosphoethanol-amine ( Figure 5). In comparison, 1D and 2D-NMR data acquired for 13 revealed the double bond of the acyl chain to be missing. Instead, an isopropyl moiety (δ H-19 1.29 ppm, δ C-20 27.7 ppm/δ H-20 0.63 ppm, δ C-20 22.2 ppm) was observed, identifying the acyl chain as iso-C 15:0 attached to the otherwise identical molecule (Table S3). Therefore, 13 was identified as 1-isopentadecanoyl-2hydroxy-sn-glycerol-3-phosphoethanolamine ( Figure 5).   Figure S3A) [15]. Furthermore, the neutral loss of 112.0885 Da (C 7 H 12 O) implied the loss of a saturated acyl group. The 1D and 2D NMR experiments confirmed this structural proposal (Table 2), showing an isopropyl group at the end of the iso-C 7:0 acyl moiety, which was attached to tyrosine via a peptide bond. Therefore, 14 was identified as N-(5-methyl)hexanoyl tyrosine ( Figure 6).  is equivalent to two additional methylene groups matching an iso-C 9:0 acyl group. Identical to NAAA 14, the acyl group of 15 was attached to tyrosine ( Figure S3B). In contrast, the fragment ion of NAAA 16 at m/z 166.0862 ([M + H] + , C 9 H 12 NO 2 + ) corresponded to one oxygen atom less and was thereby assumed to be phenylalanine ( Figure S3C). NMR analysis confirmed both structural proposals based on the MS data. NAAA 15 was determined as N-(7-methyl)octanoyl tyrosine and 16 as N-(7-methyl)octanoyl phenylalanine ( Figure 6, Table 2). Finally, advanced Marfey's analysis of NAAAs (14)(15)(16) revealed that all three compounds had been isolated as an enantiomeric mixture. The D/L-ratio was determined by UV signal integration as 1:16 for 14, 1:1.7 for 15, and 1.6:1 for 16 ( Figure S4-S6).

Antimicrobial Activity of Lipids Isolated from Bacteroidetes
The antimicrobial activity of compounds 1-4 and 12-16 isolated either from C. eiseniae DSM 22224 or Olivibacter sp. FHG000416 was determined by microbroth dilution assay against a panel of 12 indicator strains up to a test concentration of 64 µg/mL.  Table 3).

Discussion
Bacterial small molecules are of great importance for medicinal, industrial, and agricultural applications [54]. Within this class of compounds, lipids represent a structurally diverse class of metabolites with a variety of biological functions [2,26,30]. Advances in LC-MS emerged the field of lipidomics, allowing high-throughput detection and sophisticated analysis of complex lipid samples. Identification of abundant known lipid classes and structure predictions of lipids new to science are possible. However, compared with NMR techniques, LC-MS measurements that require lower amounts of sample will not provide enough data to deduce the planar structure unambiguously [39].
In the current study, we described the characterization and structure elucidation of the unknown previously determined core lipids of the Bacteroidetes genus Chitinophaga [46]. Interestingly, we observed the highest production titers of three of the four lipids in extracts of Olivibacter sp. FHG000416, a strain belonging to another Bacteroidetes genus. Therefore, compound isolation was carried out from these two strains: Olivibacter sp. FHG000416 and Chitinophaga eiseniae DSM 22,224. Two of the four lipids are identified as LPEs, with 13 being an undescribed derivative of the literature-known LPE 451 (12). The remaining ones (1 and 6), together with several further derivatives thereof, are novel lipids classified as LAAs. The literature frequently described the production of LAAs by various bacteria, with some of them being isolated and chemically fully characterized [5]. These are often mono-or dipeptide lipids containing glycine, serine, ornithine, or glycine-serine as amino acid residues amide-linked to an iso-fatty acid ester at C-3 with different degrees of unsaturation [6][7][8][9][10][11][12]. Based on MS/MS and NMR experiments, the lipids 3 and 4 were identified as new LAA glycine derivatives. The Chitinophaga core lipids 1 and 6, as well as lipids 5 and 8-11 form a LAA family with an undescribed tripeptide moiety of glycineserine-ornithine. This novel tripeptide LAA family is most closely related to the intensively studied glycine-serine dipeptide lipids 654 and 430 (2). Interestingly, with lipid 430 (2), only the de-esterified hydrolysis product of lipid 654 [34] was detected in the methanolic extracts of 23 of 25 Chitinophaga strains generated in our previous study [46]. Recent intensive studies of human pathogenic Bacteroidetes such as P. gingivalis revealed both dipeptide lipids to engage TLR2 [34]. They are involved in the development of two chronic inflammatory diseases, including periodontitis and atherosclerosis [30]. Furthermore, studies implicated lipid 654 to be involved in the development of multiple sclerosis [25]. Therefore, further studies are necessary to investigate the effects of the here described novel tripeptide LAAs in terms of immune response and whether they might also be part of the cell membrane of pathogens such as P. gingivalis.
The same applies to the new N-acyl tyrosine and phenylalanine analogs. They belong to a growing family of microbial secondary metabolites isolated from bacteria [22,23], fungi [21], or from environmental DNA expressed in heterologous hosts such as E. coli [16][17][18][19][20]. It is hypothesized that these lipids are possible signaling molecules with a wide range of biological activities from anti-cancer therapy targets to antimicrobial lipids [16][17][18][19][20][21][22][23]. With no antimicrobial activity observed, further studies are necessary to elucidate their biological function.
In addition to the structure-associated as well as the target organism-oriented antimicrobial activities of lipids, a membrane destabilization mechanism has been investigated for several decades [26,55]. In this context, a 'carpet' mechanism is suggested, which results in a detergent-like membrane permeation and/or disintegration [55,56]. Based on the reported findings, we selected a panel of 12 microorganisms to cover a wide range of possible targets for the herein described NPs. The antimicrobial profiling of the herein tested lipids revealed a frequent growth inhibitory effect towards M. catarrhalis, with lipid 1 additionally showing growth inhibiting activity against B. subtilis. The more hydrophobic cell surface of M. catarrhalis, compared with the surfaces of, e.g., E. coli and P. aeruginosa, is believed to be the reason for high accessibility of hydrophobic agents to the cell surface [57][58][59]. Therefore, we assume M. catarrhalis to be the most susceptible pathogen towards a suggested membranolytic mode of action mediated by LAAs carrying a hydrophobic lipid moiety and a short hydrophilic amino acid moiety. The mode of action needs to be confirmed in further studies.

Mass Spectrometric Analysis
For all UHPLC-QTOF-UHR-MS and MS/MS measurements, a quadrupole time-offlight spectrometer (LC-QTOF maXis II, Bruker Daltonics, Bremen, Germany) equipped with an electrospray ionization source in line with an Agilent 1290 infinity LC system (Agilent Technologies, Santa Clara, CA, USA) was used. C18 RP-UHPLC (ACQUITY UPLC BEH C18 column (130 Å, 1.7 µm, 2.1 × 100 mm)) was performed at 45 • C with the following linear gradient (A: H 2 O, 0.1% HCOOH; B: CH 3 CN, 0.1% HCOOH; flow rate: 0.6 mL/min): 0 min: 95% A; 0.30 min: 95% A; 18.00 min: 4.75% A; 18.10 min: 0% A; 22.50 min: 0% A; 22.60 min: 95% A; 25.00 min: 95% A. A 50 to 2000 m/z scan range at 1 Hz scan rate was used to acquire mass spectral data. The injection volume was set to 5 µL. MS/MS experiments were performed at 6 Hz, and the top five most intense ions in each full MS spectrum were targeted for fragmentation by higher-energy collisional dissociation at 25 eV using N 2 at 10 -2 mbar. Precursors were excluded after 2 spectra, released after 0.5 min, and reconsidered if the intensity of an excluded precursor increased by a factor of 1.5 or more. Data were analyzed using the Bruker Data Analysis 4.0 software package.
The dried culture of C. eiseniae and the 7 L culture of Olivibacter sp. were extracted with one-time culture volume MeOH for the isolation of lipid 1, lipid 430 (2), and 3, respectively. The extracts were evaporated to dryness using rotary evaporation under reduced pressure, resuspended in 3 L of 10% MeOH/H 2 O, and separately loaded onto a XAD16N column (1 L bed volume).
The 20 L fermentation of Olivibacter sp. was the starting point for the isolation of lipid 4. Due to the enlarged volume, LLE was performed as an additional purification step after MeOH extraction using ethyl acetate and water. In addition to that, the isolation procedure was highly identical for lipid 3. For semi-preparative HPLC, an adapted gradient of 55-95% CH 3 CN in water was used.

Optical Rotation
Specific rotation was determined on a digital polarimeter (P3,000, A. Krüss Optronic GmbH). Standard wavelength was the sodium D-line with 589 nm. Temperature, concentration (g/100 mL), and solvents are reported with the determined value.

Advanced Marfey's Analysis
The absolute configuration of all amino acids was determined by derivatization using Marfey's reagent [50]. Stock solutions of amino acid standards (50 mM in H 2 O), NaHCO 3 (1 M in H 2 O), and L-FDVA (70 mM in acetone) were prepared. Commercially available standards were derivatized using molar ratios of amino acid to FDVA and NaHCO 3 (1/1.4/8). After stirring at 40 • C for 3 h, 1 M HCl was added to obtain a final concentration of 170 mM to end the reaction. Samples were subsequently evaporated to dryness and dissolved in DMSO (final concentration 50 mM). L-and D-amino acids were analyzed separately using C18 RP-UHPLC-MS with the standard gradient (for details see Section 4.2) at a flow rate of 0.6 mL/min. Total hydrolysis of compounds 1, lipid 430 (2), and 14-16 was carried out by dissolving 250 µg of each compound in 6 M deutero-hydrochloric acid (DCl in D 2 O) and stirring for 7 h at 160 • C. The sample was subsequently evaporated to dryness. Samples were dissolved in 100 µL H 2 O, derivatized, and analyzed using the same parameters as described before.

Minimal Inhibitory Concentration (MIC)
Microbroth dilution assays were performed in 96-well plates to determine the minimum inhibitory concentrations (MIC) of purified compounds dissolved in DMSO and were tested in triplicate following EUCAST instructions with minor adaptions [63,64]. A cell concentration of 5 × 10 5 cells/mL was adjusted for all bacteria from an overnight culture (37 • C, 180 rpm) in cation-adjusted Mueller-Hinton II medium (BD). All tested organisms are summarized in Table 3. Dilution series of rifampicin, tetracycline, and gentamicin were used as control antibiotics (64-0.03 µg/mL) to ensure that concentrations achieved a range of effects from none to complete growth inhibition of the test strain. Negative controls were cell suspensions without test sample or antibiotic control. The turbidity was measured with a microplate spectrophotometer at 600 nm (LUMIstar Omega BMG Labtech) to assess cell growth after overnight incubation (18 h, 37 • C, 180 rpm, 85% rH).
Mycobacterium smegmatis ATCC 607 was grown in brain-heart infusion broth supplemented with Tween 80 (1.0% v/v) at 37 • C and 180 rpm for 48 h before the cell density was adjusted in cation-adjusted Mueller-Hinton II medium. The gentamicin control was replaced with isoniazid. Assay read out was done by cell viability assessment after 48 h (37 • C, 180 rpm, 85% rH) by ATP quantification (BacTiter-Glo, Promega), according to the manufacturer's instructions.
Candida albicans FH2173 was incubated at 28 • C for 48 h before diluting to 1 × 10 6 cells/mL in cation-adjusted Mueller-Hinton II medium. Assays were incubated at 37 • C for 48 h with nystatin as positive control and were evaluated by ATP quantification (BacTiter-Glo, Promega).
Pre-and main cultures of Micrococcus luteus DSM 20030 and Listeria monocytogenes DSM 20600 were incubated for two days, and the assay readout was done by ATP quantification, as described before.  Table S1: Overview of all lipids with their molecular formula, predicted and found masses within the metabolomics data of our previous study, and in how many of the investigated 25 Chitinophaga metabolomes each lipid is present,