Bacterial cultivation and extraction of metabolites
A. muciniphila BAA-835 was inoculated in 3 ml of BHI medium in a 5 ml Falcon tube and incubated under anaerobic conditions at 37 °C for 4 days. Then, 3 ml of the culture was used to inoculate 1 l of BHI medium with 1.5 g of mucin from porcine stomach (Sigma-Aldrich), in a 1 l Pyrex storage bottle (16 bottles × 1 l each, total volume 16 l) and the cultures were incubated for 12 days under anaerobic conditions at 37 °C. After 12 days of static growth, bacterial cultures were centrifuged to separate cell pellets and supernatants (8,000 r.p.m. for 30 min). The cell pellets were extracted with chloroform and methanol (1:1) by stirring for 24 h at room temperature. The solvent mixture was filtered through Whatman qualitative filter paper (grade 3, circle, diameter 125 mm) and dried under vacuum.
For the extraction of supernatants, 100 g of hydrophobic resin mixture (Amberlite XAD4HP and XAD7HP, 20–60 mesh) was added directly to spent media to allow secreted metabolites to adsorb to the resins. Then, the resin mixture containing bacterial metabolites was washed with acetone and methanol (1:1) and stirred for 24 h at room temperature. The solvent mixture with resin mixture was filtered through Whatman qualitative filter paper (grade 3, circle, diameter 125 mm) and concentrated on a rotary evaporator. The cultivation and extraction procedures were repeated eight times (total culture volume, 128 l), yielding 4 g of dry extract from the cell pellets and 15 g of crude extract from the supernatants.
Bioassay-guided fractionation, purification and identification of a15:0-i15:0 PE
The crude extract from the cell pellets (4 g) was dissolved in chloroform and fractionated by normal-phase chromatography using seven different solvent systems (A, 100% hexane; B, 100% chloroform; C, 100% ethyl acetate; D, 75% ethyl acetate/25% methanol; E, 90% acetone/10% methanol; F, 50% methanol/50% dichloromethane; and G, 100% methanol) with a silica column (Teledyne Isco, RediSep RF Gold Silica 12 g). The pro-inflammatory activity was highly detected in fractions F and G. The mixture of fractions F and G (120 mg and 210 mg, respectively, and 8.3% of total yield) was then subjected to reversed-phase semi-preparative high-performance liquid chromatography (HPLC) (Luna C8 (2), 250 × 10 mm, 5 µm) using the following gradient solvent system: 10% methanol/90% water isocratic for 10 min; gradient to 30% methanol/70% water for 10 min; then 30% methanol/70% water to 90% methanol/10% water for 20 min, 90% methanol isocratic for 10 min, gradient to 100% methanol for 25 min; flow rate, 2 ml min−1). Fractions were collected every 1 min between 5 min and 75 min, generating 70 fractions. Fractions able to stimulate pro-inflammatory cytokine production from mBMDCs were combined and identified as bacterial PE with BCFAs (22 mg, yield = 0.55%). An essentially pure compound, later identified as a15:0-i15:0 PE, was acquired at a retention time of 63 min (14 mg, yield = 0.35%).
The crude extract from the supernatants (15 g) was dissolved in methanol and filtered through a syringe filter (polytetrafluoroethylene (PTFE), 0.2 µm). The filtered extract was directly injected onto a reversed-phase preparative HPLC column (Luna C18 (2), 250 × 21.2 mm, 5 µm) with a gradient mobile solution (30% methanol/70% water to 100% methanol for 30 min, 100% methanol isocratic for 30 min; flow rate, 10 ml min−1). Fractions were collected every 2 min from 5 min to 55 min, generating 25 fractions. Fractions able to stimulate pro-inflammatory cytokine production from mBMDCs were collected at 50 min (23.5 mg, 0.16%) and further purified as described above, resulting in additional a15:0-i15:0 (1.2 mg, yield = 0.008%). Overall, 15.2 mg of a15:0-i15:0 was isolated from 19.0 g of crude extract (0.08%).
The structure of a15:0-i15:0 PE was identified by the comprehensive analysis of 1H, 13C and two-dimensional (2D) NMR spectroscopic data (Extended Data Table 1).
NMR spectroscopy
All 1H NMR spectra were acquired at 500 MHz at 30 °C, and chemical shifts are represented on a δ (beta) scale. Residual protium in the NMR solvent (CDCl3, δ 7.26) was used to reference chemical shifts. Data are represented as follows: assignment, chemical shift, integration, multiplicity (s, singlet; d, doublet; t, triplet; q, quartet; m, multiplet; br, broad) and coupling constant in hertz. All 13C NMR spectra were obtained at 125 MHz at 30 °C and chemical shifts are represented on a δ scale. The carbon resonances of the NMR solvent (CDCl3, δ 77.17) were used to reference chemical shifts. Full assignment of protons and carbons were completed on the basis of the following 2D NMR spectroscopy experiments: gradient 1H–1H correlation spectroscopy, gradient 1H–13C heteronuclear single quantum coherence, gradient 1H–13C heteronuclear multiple bond connectivity. Mnova v.14.2.0 was used to analyse NMR data of natural and synthetic compounds.
High-resolution mass spectrometry for a15:0-i15:0 PE and other family members
High-resolution mass spectrometry data were collected using Agilent MassHunter Work Station LC/MS Data Acquisition 10.1 and Agilent LC-QTOF Mass Spectrometer 6530 equipped with a 1290 uHPLC system and electrospray ionization detector scanning from m/z 50 to 3,200. Then 5 μl aliquots of a15:0-i15:0 PE and its family members were injected into a reversed-phase analytical column (Luna C8: 100 × 2.1 mm, 5 μm) using a gradient solvent system with 0.1% formic acid (10% methanol/water to 90% methanol/water for 10 min, 90% methanol/water isocratic for 10 min, then gradient to 100% for 10 min; flow rate, 0.3 ml). Agilent MassHunter Qualitative Analysis B.07.00 software was used to analyse the data.
FA methyl esterification and GC–MS analysis of A. muciniphila PE
A 0.1 mg sample of both a15:0-i15:0 PE and complete A. muciniphila PE were dissolved in 200 μl of methanol, and 1.4 mg of sodium methoxide was added to prepare a 0.5 mol l−1 sodium methoxide solution. The reaction mixture was stirred at room temperature for 3 h then quenched by addition of 1N HCl. The methanolysis products were dried under vacuum and extracted with ethyl acetate and water (300 μl, v/v = 2:1). The water layers were removed, and each of the ethyl acetate layers containing FA methyl esters (FAME) were injected into a gas chromatograph (GC, Agilent MassHunter GC/MS Acquisition B.07.05.2479) combined with a HP-5 ms Ultra Inert column (0.25 mm × 30 m). The temperature of the injector and the detector in the GC was maintained at 150 °C. During analysis, the temperature of the GC column was controlled (150 °C for 3 min, 150–250 °C at 6 °C min−1 and 250 °C for 3 min). The FAME derivatives of a15:0-i15:0 PE were composed of i15:0 and a15:0 (1:1 ratio) having retention times at 10.2 min and 9.7 min, respectively. The gas chromatography–mass spectrometry (GC–MS) analysis of FAME derivatives of AmPE displayed i14:0 (15.7%), n14:0 (2.7%), a15:0 (51.7%), i15:0 (23.6%), a16:0 (0.6%), i16:0 (1.8%), a17:0 (1.7%) and a18:0 (2.2%), having retention times at 8.0, 8.6, 9.7, 10.2, 11.3, 11.9, 13.4 and 15.0 min, respectively (Fig. 1d). Agilent MassHunter Qualitative Analysis B.07.00 software was used to analyse GC–MS data.
O-deacylation for determination of a15:0 connected to sn-1
A 5 mg sample of a15:0-i15:0 PE was prepared and lyophilized for 24 h. A 1 mg ml−1 of NaOMe solution was prepared, and the mixture was dissolved in 500 μl of NaOMe solution at room temperature. The solution was stirred under argon for 30 min. After 30 min, the reaction was quenched by addition of 1N HCl and dried under vacuum. The O-deacylated product, a15:0 PE, was purified by reversed-phase HPLC (Luna C8 (2): 250 × 10 mm, 5 μm) with an isocratic solvent system (45% acetonitrile/water over 30 min, ultraviolet 210 nm detection, flow rate 2 ml min−1). The O-deacylated product (1.8 mg) was eluted at 12.5 min, and its structure was determined by one-dimensional and/or 2D NMR spectroscopy (Extended Data Table 2) and by low-resolution electrospray ionization mass spectrometry (ESI-MS) ([M+H]+m/z at 440; molecular formula, C20H43NO7P).
Amino-acid feeding experiment
A volume of 5 ml of A. muciniphila BAA-835 grown in BHI was inoculated into three 1 l bottles of M9 medium supplemented with 1.5 g of mucin from porcine stomach (Sigma-Aldric) and either 1 mmol l−1 of l-leucine, l-isoleucine or l-leucine/l-isoleucine mixture (1:1 ratio) or nothing as a control. The cultures were grown under anaerobic conditions at 37 °C for 12 days. The cell pellets from these cultures were centrifuged and extracted with 40 ml of chloroform and methanol (1:1). The extract was dried under vacuum and dissolved in dimethyl sulfoxide at a 10 mg ml−1 concentration and tested for activity in the mBMDC cytokine assay. Statistical significances were determined using an unpaired two-tailed Student’s t-test.
a15:0-i15:0 PE biosynthetic gene identification and analysis
Sequence comparison and analysis of the a15:0-i15:0 PE biosynthetic pathway to the previously reported BCFA biosynthetic pathway and de novo biosynthetic pathway of leucine, isoleucine and valine were performed using blastp (NCBI RefSeq database, updated 8 September 2015), Kyoto Encyclopaedia of Genes and Genomes and Geneious v.11.1.4 for pairwise sequence alignments that were previously reported. The accession number for the genes used in this analysis is CP001071.1.
Total synthesis for small library of PEs
The total synthesis of a15:0-i15:0 PE, i15:0-a15:0 PE, a15:0-a15:0 PE, i15:0-i15:0 PE and n15:0-n15:0 PE was performed by previously reported methods38,39,40.
Animal and human-cell studies
Mouse experimental procedures complied with all relevant ethical regulations and were conducted according to protocol 2003N000158 approved by the Institutional Animal Care and Use Committee at Massachusetts General Hospital. Appropriate sample sizes were estimated based on the effect size and variance of cytokine measurements in myeloid cells stimulated with canonical TLR ligands. In all mouse experiments, animals were allocated to experimental groups based on genotype and/or age and sex matched. Male or female wild-type, TLR2−/− or TLR4−/− C57BL/6 mice at least 3–4-weeks old and preferably 7–12 weeks of age were used. Mice were housed with a 12 h light or dark cycle at an ambient temperature of between 18 °C and 24 °C and a relative humidity of between 30% and 70%.
Human monocytes were isolated from buffy coats collected from healthy donors at the Blood Donor Center at Massachusetts General Hospital in compliance with all relevant ethical regulations and according to protocol 2018P001504 approved by the Mass General Brigham Institutional Review Board. Donors provided informed written consent.
mBMDC cytokine assays
These assays were done as described previously8. In brief, femurs and tibias were collected from male or female wild-type, TLR2−/− or TLR4−/− C57BL/6 mice that were at least 3–4-weeks old and preferably 7–12 weeks of age. The bone marrow was pushed from the bones using a needle and syringe of complete Dulbecco’s Modified Eagle Medium (DMEM) supplemented with Gibco GlutaMAX Supplement (35050061), Gibco penicillin-streptomycin (15140122) and 10% heat-inactivated foetal bovine serum (FBS), and strained through a 70 µm nylon filter. The collected bone marrow was then centrifuged, and red blood cells were lysed using Invitrogen eBioscience 1X RBC Lysis Buffer (00-4333-57). The cells were then centrifuged and strained through a 70 µm nylon filter again and resuspended in complete DMEM. Cells were counted and then plated at approximately 5 million cells per plate with approximately 20–40 ng ml−1 recombinant murine (Rm) GM-CSF (PreProTech 315-03). They were allowed to grow for 7 days, sometimes with additional feeding of 20–40 ng ml−1 RmGM-CSF on day 3. The resulting mBMDCs were then scraped and counted again. They were plated in 96-well tissue culture-treated microplates (Corning CLS3599) from 50,000–90,000 cells per well and allowed to adhere for at least 3 h. The cells were then treated with chromatographic fractions or purified compounds at a final concentration of 50 µg ml−1, a final concentration of LPS (InvivoGen tlrl-b5lps) of 3–625 ng ml−1 or a final concentration of Pam3CSK4 (InvivoGen tlrl-pms) of 0.250–1.562 µg ml−1 and incubated overnight. The following morning, supernatant was removed and an enzyme-linked immunosorbent assay (ELISA) was performed to measure TNFα using an Invitrogen Mouse ELISA kit (88-7324-77) per the manufacturer’s instructions. Gen5 v.3.03 or SoftMax Pro v.6.2.1 (SpectraMax, Molecular Devices) was used to analyse ELISA plates. For cytokine detection using flow cytometry, we used the cytometric bead array mouse inflammation kit from BD Biosciences (552364) per the manufacturer’s instructions. Data were collected with NovoExpress v.1.4.1 and analysed data using FlowJo v.10.7.
Peripheral blood mononuclear cell cytokine assay
Peripheral blood mononuclear cells (PBMCs) were enriched for monocytes using the RosetteSepHuman Monocyte Enrichment Cocktail (STEMCELL Technologies, catalogue no. 15028). In brief, buffy coats were incubated with monocyte enrichment cocktail for 20 min at room temperature while rocking. They were then diluted with 1X phosphate-buffered saline (PBS) and layered over the Ficoll-Paque PLUS medium (GE Healthcare, catalogue no. 17-1440-02) and centrifuged for 20 min at 1,200g. Enriched monocytes were collected and cultured with chromatographic fractions or purified compounds at 50 µg ml−1 in DMEM media containing 10% FBS and 1% penicillin-streptomycin. LPS and Pam3CSK4 at a final concentration of 100 ng ml−1 were used as controls. After overnight incubation, supernatant was collected and analysed for IL-6, IL-10, IL-12/IL-23p40 and TNFα cytokines using Human Flex Set Kits (BD CBA, catalogue nos. 558276, 558274, 560154 and 560112).
RNA sequencing
Monocytes were isolated from PBMCs as described previously41. Bulk RNA sequencing libraries were prepared using SmartSeq2. Libraries were sequenced on a NextSeq (Illumina). FastQC v.0.11.5 and MultiQC v.1.8 were used to confirm the quality of the sequenced libraries42,43. Next, kallisto v.0.46.1 was used with a GRCh38 reference to generate the counts of reads mapped to each gene44,45. The matrix of counts was used for the calculation of counts per million (CPM) values, and the generated CPM matrix was treated with log2(CPM + 1) to obtain a log expression matrix. A gene with a CPM value greater than 1 was considered as expressed. Samples obtained after the above steps were then used to detect differentially expressed genes via EdgeR v.3.35.1 (ref. 44). The lists of differentially expressed genes were generated from likelihood ratio tests based on the generative linear model framework, following the prerequisite gene filtering, normalization and dispersion estimation steps of the software.
CRISPR targeting
PBMCs were isolated from buffy coats using Sepmate tubes (STEMCELL Technologies) and ammonium–chloride–potassium lysis buffer following the manufacturer’s protocol. Human monocytes were harvested from PBMCs by negative selection using RosetteSep human Monocyte Enriched Cocktail (STEMCELL Technologies) according to the manufacturer’s protocol. Alt-R sgRNAs were purchased from IDT and reconstituted to 100 µmol l−1 with Nuclease-Free Duplex Buffer (IDT). In a sterile polymerase chain reaction strip, the sgRNAs were mixed with Cas9 (IDT, Alt-R S.p. Cas9 Nuclease V3) at a molar ratio of 2:1 (2 µl sgRNA at 100 µmol l−1 + 2 µl Cas9 at 5 mg ml−1) for each reaction and incubated at room temperature for over 20 min. Monocytes were washed twice with 5 ml of PBS and counted. Then 2 × 106 cells per reaction were resuspended in 16 µl of P3 primary nucleofection solution (Lonza). The 16 µl of cells in P3 buffer was added to each Cas9–ribonucleoprotein complex. The cell–ribonucleoprotein mix was then immediately loaded into the supplied nucleofector cassette strip (Lonza) and nucleofected using 4D-Nucleofector with CM-137 programme. Then 180 µl of prewarmed medium was immediately added into each cassette well. A volume of 1 × 105 cells was seeded into a 96-well plate with medium (RPMI-1640 with 10% FBS, 2 mmol l−1 Glutamax, 55 µmol l−1 beta-mercaptoethanol, 100 U ml−1 penicillin, 100 µg ml−1 streptomycin, GM-CSF 800 U ml−1 and IL-4 500 U ml−1). The medium was changed every 2–3 days. At day 5, MDDCs were stimulated with 10 µg ml−1 of Akkermansia lipids, 100 ng ml−1 of Pam3CSK4, 100 ng ml−1 of FSL-1 or 100 ng ml−1 of LPS for 18 h or as indicated. Cell supernatants were collected for human TNFα measurements by ELISA (Invitrogen) following the manufacturer’s protocol. SoftMax Pro v.6.2.1 (SpectraMax, Molecular Devices) was used to analyse ELISA plates. The sgRNA sequences used were as follows:
Human TLR1: GGTCTTAGGAGAGACTTATG
Human TLR2: GACCGCAATGGTATCTGCAA
Human TLR6: ATTCATTTCCGTCGGAGAAC
TLR2–TLR1–a15:0-i15:0 PE complex modelling
Modelling of the a15:0-i15:0 PE ligand complex was based on the crystal structure of the TLR2–TLR1–Pam3CSK4 complex from the Protein Data Bank (PDB ID 2z7x)28. The Pam3CSK4 ligand was removed from the crystal structure coordinates, and an a15:0-i15:0 PE ligand was prepared using Lidia and AceDRG in Coot v.0.9 (refs. 46,47). The a15:0-i15:0 PE ligand placement in the ligand-binding pockets of TLR2 and TLR1 was guided by the electron density belonging to the acyl chains of the Pam3CSK4 ligand in the crystal structure. Structural figures and videos were generated using ChimeraX v.1.0 (ref. 48). Structural biology software was compiled and configured by SBGrid consortium49.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.
