Novel CB1-ligands maintain homeostasis of the endocannabinoid system in ω3- and ω6-long-chain-PUFA deficiency[S]

Mammalian ω3- and ω6-PUFAs are synthesized from essential fatty acids (EFAs) or supplied by the diet. PUFAs are constitutive elements of membrane architecture and precursors of lipid signaling molecules. EFAs and long-chain (LC)-PUFAs are precursors in the synthesis of endocannabinoid ligands of Gi/o protein-coupled cannabinoid receptor (CB)1 and CB2 in the endocannabinoid system, which critically regulate energy homeostasis as the metabolic signaling system in hypothalamic neuronal circuits and behavioral parameters. We utilized the auxotrophic fatty acid desaturase 2-deficient (fads2−/−) mouse, deficient in LC-PUFA synthesis, to follow the age-dependent dynamics of the PUFA pattern in the CNS-phospholipidome in unbiased dietary studies of three cohorts on sustained LC-PUFA-free ω6-arachidonic acid- and DHA-supplemented diets and their impact on the precursor pool of CB1 ligands. We discovered the transformation of eicosa-all cis-5,11,14-trienoic acid, uncommon in mammalian lipidomes, into two novel endocannabinoids, 20:35,11,14-ethanolamide and 2-20:35,11,14-glycerol. Their function as ligands of CB1 has been characterized in HEK293 cells. Labeling experiments excluded Δ8-desaturase activity and proved the position specificity of FADS2. The fads2−/− mutant might serve as an unbiased model in vivo in the development of novel CB1 agonists and antagonists.

and 2-arachidonoyl-glycerol (2-AG) are the most dominant ligands of cannabinoid receptor (CB)1 and CB2. The endocannabinoid system (ECS) (7)(8)(9)(10) consists of endocannabinoids, receptors CB1 and CB2, and associated anabolic and catabolic enzymes (11)(12)(13), and modulates the orexinergic inputs into the neuronal network in selective regions of the CNS, sensing nutrient availability for maintaining body energy homeostasis (14). Unlike the intensively studied structure-function relationship of CB1 and CB2 in the ECS of the CNS [for review see (15)], the role of the dietary supply of PUFAs and their transformation to lipophilic CB1 ligands has remained elusive.
In this study, we utilized the auxotrophic fads2 / mouse mutant to explore the impact of a) EFAs in the nd-fads2 / mouse, b) 6-AA in the AA-fads2 / mouse, and c) 3-DHA in the DHA-fads2 / mouse on the PUFA pattern in diacylglycerol structures of their phospholipidome and the derived endocannabinoid metabolomes of brain and the resulting physiological states of the system.

Fads2
/ mice The fads2 / mouse line was developed in this laboratory (16) and, after 10 backcrossings, maintained on a C57Bl/6N background. WT mice were obtained from heterozygous fads2 +/ ×fads2 +/ crossings. All mice were genotyped by PCR analysis of tail DNA. Cohorts of gender-and weight-matched WT and fads2 / mice, of ages indicated in the experiments, were used in this study. Animal protocols followed the principles and practices outlined in the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and breeding and tests were done with permission of the local authorities (LANUV NRW). The animal studies reported in this work followed the ARRIVE-Guidelines (19). All animals were kept under specific pathogen-free conditions. The light/dark cycle was 12/12 h; the animals had free access to water and a regular diet (nd), 6-20:4 (AA)-supplemented diet, or 3-22:6 (DHA)-supplemented diet (Altromin diet #1310; Altromin, Lage, Germany). Table 1 summarizes the GC/MS analysis of the fatty acid composition of the diets used for the transformation of the nd-fads2 / into the AA-and DHA-fads2 / mouse lines. Altromin diet #1310 was used as the basic diet and contains the two EFAs, 18:2 and -18:3, to prohibit EFA deficiency. AA was supplemented as ARASCO and DHA as DHASCO triglycerides, with 50% 20:4 (AA) or 22:6 (DHA) as single LC-PUFAs, respectively, in the nd. Heterozygous foster mothers and colonies of +/+ and / mice were maintained on the respective diets throughout their lifetime.

Lipidome analysis
Total lipids were extracted from brain or serum of WT and fads2 / male and female mice, homogenized in an Ultraturrax in 10 vol of chloroform/methanol (C/M) 2:1 (v/v) and re-extracted with C/M 1:1 (v/v) and C/M 1:2 (v/v) for 1 h each at 37°C under a stream of nitrogen. The combined extracts of total lipids were dissolved in C/M 2:1 (v/v), washed with 2 M KCl and water, and taken to dryness in a stream of nitrogen. Phospholipids were separated in the solvent system chloroform/ethanol/triethylamine/water 60/70/70/14 (v/v/v/v), using HPTLC plates (Merck, Germany). Bands were visualized by primuline fluorescence (0.2% in 80% acetone). Lipid bands were collected on fritted glass filters and eluted with C/M 2:1 (v/v) into Sovirel tubes and concentrated under N 2 for analysis by MS. For GC/MS analysis, fatty acids were esterified with 5% HCl-methanol at 80°C for 1 h. One volume of water was added and fatty acid methyl esters (FAMEs) extracted with hexane and concentrated under nitrogen.

Synthesis of [D6
Intermediates and labeled and unlabeled end products were controlled by GC/MS and NMR.

Endocannabinoid analysis
Endocannabinoids were isolated following a modified previously described method (21). In brief, internal standards of D8-20:4 AEA, D8-20:4 AG, and D4-18:2 linoleoyl-ethanolamide (LEA) were supplied to brain and serum samples of WT and fads2 / female mice and extracted with acetone/PBS 3:1 in an Ultra-Turrax homogenizer. Samples were centrifuged (10,000 g, 10 min, 4°C). Supernatants were transferred to silanized tubes and evaporated to dryness under nitrogen. PBS (100 l) and C/M (300 l) 2:1 were added to the residue, vortexed for 15 s, and the two phases separated by centrifugation. The bottom phase was evaporated and reconstituted in 100 l of methanol/water 3:1.
High resolution full scan MS measurements were accomplished using an Agilent Technologies (Waldbronn, Germany) HPLC coupled to an AB Sciex (Darmstadt, Germany) TripleTOF 5600 mass spectrometer. The system was equipped with a Thermo Fisher Scientific (Dreieich, Germany) Accucore C8 (2.6 m particle size, 50 × 3 mm) analytical column, and mobile phases consisted of aqueous formic acid 0.2% (pH 2) (solvent A) and acetonitrile as organic modifier (solvent B) for optimal ESI conditions. Thereby, the gradient was held for 0.5 min and then decreased from 50% A to 0% A with 0.325 ml/min within 6 min. Afterwards, the column was washed for 1 min with 100% B. The subsequent re-equilibration time was 3 min. The mass spectrometer was calibrated frequently (after 10 injections) via the Duo Turbo-V-Ion source by a calibrant delivery system containing the manufacturer's calibrants for positive ionization. The nitrogen for the ion source as well as collision gas supply was delivered by the nitrogen generator (CMC, Eschborn, Germany). Product ion experiments were acquired by isolating the respective [M+H] + precursor ions in the quadrupole (unit resolution) and performing collision-induced fragmentation in the collision cell.

Cell culture, transfection, and incubation experiments
WT and fads2 / mouse embryonal fibroblasts (MEFs) were prepared according to an established protocol (22). MEFs and HEK293 cells were grown in Dulbecco's modified Eagle's medium (Seromed) supplemented with 10% fetal calf serum (Life Technologies, Inc.), 2 mmol glutamine, 1 mM sodium pyruvate (Biochrom), 100 units/ml penicillin, and 100 g/ml streptomycin in a humidified incubator at 37°C in a 5% CO 2 atmosphere. ] i were measured using the Fluo-4AM (Thermo Fisher Scientific) calcium indicator as described in (23). Briefly, cells were loaded with 5 M of Fluo-4AM dissolved in sterile imaging buffer [125 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1.2 Mm MgSO 4 ·7H 2 O, 25 mM HEPES (pH 7.4)] for 20 min at room temperature and subsequently washed with 2× 1 ml of imaging buffer. Each measurement consisted of a series of 500 images recorded at a rate of 0.254 s per frame (512 pixels 2 ) for a total of 127 s via the FV-ASW 1.7 software on an Olympus FV1000 confocal microscope (488 nm excitation at 0.3% intensity, >505 nm emission). Stimulation of CB1 was measured by the fluorescence intensity evoked by Ca 2+ release after ligand addition 5-10 s after the initiation of an image series recording.
Individual cells were marked as regions of interest (ROIs), and the average fluorescence was measured for each ROI. Background of each image was subtracted from each pixel of the average fluorescence (F) of a cell-free area. To account for the fluorescence of the expression marker, EGFP, a short series of 10 frames was taken immediately prior to Fluo-4AM loading and comparison of the average fluorescence in 10 ROIs before and after loading was used to calculate the increase in fluorescence caused by Fluo-4AM loading. For the analysis of CB1 agonistinduced Ca 2+ transients, the fluorescent signals emanating from each of 10 ROIs (i.e., 10 cells) were evaluated for each image series via the formula F/F 0 , where F 0 is the average fluorescence intensity of the first five frames taken before ligand addition and F is F  F 0 .

Real-time PCR
RNA was isolated from WT and fads2 / male and female brains of littermates using Trizol (Invitrogen). Total RNA (10 g) was reverse-transcribed using a Transcriptase kit (Life Technologies). Primer pairs used in quantitative PCR reactions are listed in Table 2. Hgprt was used as internal standard. Quantitative PCR reactions were performed with the ABI Prism 7900HT employing a 96-well format and Fast SYBR Green Master Mix (Applied Biosystems) following the manufacturer's protocol. Data analysis was performed using the 2 Ct method.

Protein analysis by Western blotting
Freshly dissected brains of WT and fads2 / male and female mice were mechanically homogenized in lysate buffer containing protease inhibitor cocktail (Complete; Roche). Protein concentrations were measured using a Pierce BCA protein assay kit (Thermo Scientific).
Sections were washed with PBS/0.05% Tween20, incubated with Cy3-conjugated secondary IgG antibody (Jackson Immuno Research) for 1 h at 37°C, washed with PBS/0.05% Tween20, and immunostained with Affinity purified rabbit polyclonal or monoclonal antibodies and HRP with DAB substrate (Roche #1718096) in the peroxidase reaction. A Zeiss microscope, Axio Imager.M1, and the AxioVision imaging software (RRID:SCR_002677) were used for fluorescence microscopy and a Slide Scanner Leica SCN400 and the Aperio ImageScope software (RRID:SCR_014311) for peroxidase microscopy.

Thermal nociception test
The hot plate test measuring thermal nociception was carried out following established procedures (24). The plate temperature was maintained at 54-55°C. Cut-off time was set at 60 s. WT and fads2 / female mice were immediately removed from the hot plate when the pain threshold was reached.

Statistical analysis
Results are expressed as mean ± SEM. Statistical analysis of differences between individual experimental groups was performed using GraphPad QuickCalcs. Sample sizes are indicated in the figure legends. Student's unpaired two-tailed t-test was used. P-values of 0.05*, 0.01**, and 0.001*** were considered significant.
GC/MS analysis of the fatty acid pattern of phospholipids revealed regression of 6-AA and 3-DHA concentrations in all peripheral tissues of auxotrophic nd-fads2 / siblings within 4 months, here exemplified in total lipid extracts of liver (supplemental Fig. S1B). The liver lipidome of nd-fads2 / mice contained EFAs, but was deprived of 6and 3-LC-PUFAs. Instead, 6-sciadonic acid was incorporated as surrogate of PUFAs in the diacylglycerol-backbones of all phospholipids. Also, the PUFA pattern of serum phospholipids of nd-, 6-AA-, and 3-DHA-WT and -fads2 / mice (supplemental Fig. S1C-E) showed absence of LC-PUFAs in nd-fads2 / mice (supplemental Fig. S1C).
We next focused on the PUFA pattern of the brain phospholipidomes of adult (4 months) nd-, 6-AA-, and 3-DHA-WT and -fads2 / mice, as critical sources of PUFA substrates for on-demand endocannabinoid synthesis. Phospholipid classes were separated by HPTLC, and their FAME substituents identified and quantitated by GC/MS ( Fig. 1A-C).
Following this surprising observation, we measured the kinetics of the retarded PUFA exchange in the brains of 8-, 12-, 16-, and 30-week-old nd-WT and nd-fads2 / mice, taking PE species as paradigm. 6-Sciadonic acid increased to a final concentration of 10% of total fatty acids within 12 weeks, 6-AA and 3-DHA inversely decreased to 10% after 30 weeks and persisted at this level during the lifespan (Fig. 1D, E).

Regional distribution of CB1 and OX1R expression in fads2
/ mice Regional expression of CB1 and OX1R, representing the ECS and orexinergic system of brains of nd-WT and -fads2 / mice, was visualized in coronal sections by IHC, using anti-CB1 and anti-OX1R antibodies and anti-HRPlabeled secondary antibody (Fig. 5A-D). Neurons in hypothalamus, visual cortex 2 mediolateral, and dentate gyrus showed pronounced CB1 (Fig. 5A, B) and OX1R expression in posterior hypothalamus of nd-fads2 / mice compared with nd-WT mice (Fig. 5C, D).
IHC signal intensities in images of fluorescence-stained coronal sections recorded under identical parameters indicated enhanced expression of CB1 in nd-fads2 / visual cortex, FAAH in the nd-fads2 / CA1 region, and OX1R in the CA1 region of 6-AA-fads2 / brain. Fluorescence intensities of CB1 in the visual cortex and of CA3 and FAAH in the CA1 and CA3 regions in 3-DHA-WT brain were strikingly higher than in 3-DHA-fads2 / brain (Fig.  5E-G).

Activator function of SEA and 2-SG in cb1-expressing HEK293 cells
Mock-and cb1-transfected HEK293 cells were characterized as described (supplemental Fig. S4). Semi-confluent cultures were exposed to SEA and 2-SG and compared with AEA and 2-AG. Their function in activating the G i/o protein-coupled CB1 was tested by recording changes in intensities of intracellular [Ca 2+ ] transient-induced Fluo-4 fluorescence.
Comparable time-dependent corrected fluorescent intensities (F/F 0 ) induced by AEA and 2-AG and the novel SEA and 2-SG ligands were recorded in single cb1-transfected HEK293 cells (Fig. 6A-D).
We tested nociception in the three mouse cohorts in the hot plate test. Prolonged latency was observed in nd-fads2 / and 6-AA-fads2 / mice compared with their WT Novel CB1 ligands in PUFA deficiency 1405 controls (Fig. 7A). Thermo-sensitivity of 3-DHA-fads2 / mice was significantly higher compared with 3-DHA-WT mice. The most notable observation was prolonged latency in 6-AA-WT and AA-fads2 / mice to almost twice that of nd-WT and nd-fads2 / and 3-DHA-WT and DHA-fads2 (Fig. 7A).
Rectal temperatures of WT mice and fads2 / mice in the three cohorts remained unchanged (Fig. 7B).

DISCUSSION
The comparative analysis of the turnover of PUFAs in the phospholipidomes of CNS and extraneuronal tissues of WT and fads2 / mice revealed the postnatal systemic complete depletion of LC-PUFAs in peripheral tissues, unlike the loss of LC-PUFA substituents and inverse substitution by 6-sciadonic acid in the diacylglycerol backbone of the phospholipidome of brain, which commenced after weaning and proceeded to a low but constant level and persisted during the lifespan.
The systemic absence of 6-AA in the adult fads2 / mouse precludes the desaturation of 6-sciadonic acid to 6-AA by a 8-desaturase as enzyme entity. Tracing experiments with fads2 / MEFs, using labeled 6-linoleic acid and 6-sciadonic acid as substrates, detected no synthesis of labeled 6-AA. Absence of the transformation of 6sciadonic acid to 6-AA in fads2-overexpressing HEK293 cells excluded 8-desaturase activity. This underlines the position specificity of the enzyme of FADS2.
This suggests dietary 6-linoleic acid supply to be sufficient, maintaining homeostasis of the mammalian ECS.
Gene and protein expression of key players in the ECS and the orexinergic system in brains of nd-, AA-, and DHA-fads2 / cohorts indicated remarkable changes in cannabimimetic effects of the novel endocannabinoids and their function in their connectivity to the orexinergic system.
The synthesis of endocannabinoid ligands of CB1 in the ECS of the CNS in the fads2 / mouse depends on the dietary supply of EFAs and preformed LC-PUFAs. The PUFA ratio in current Western diet is regarded as a critical nutritional parameter for numerous cardiovascular, metabolic, and neurodegenerative diseases and is regarded as a putative epigenetic factor (1,3,4).
Unexpectedly, the phospholipidome of brains of adult nd-fads2 / mice contained persisting 3and 6-PUFA substituents (Fig. 1A). Phospholipid class-specific LC-PUFA depletion followed a linear regression starting after weaning and reached low but constant levels within 6-7 months, which persist during the lifespan (Fig. 1D,E).
The kinetics clearly indicated the caveats of short-time feeding experiments, aiming at manipulating the lipid bilayer architecture of neuronal membrane systems as microenvironment of functionally diverse integral membrane proteins.

Structure-function relationship of SEA and 2-SG as CB1 ligands
A remarkable neuronal response to LC-PUFA deficiency was the synthesis of two novel endocannabinoids, which were identified as SEA and 2-SG and quantified by LC/ MS/MS ( Fig. 3; supplemental Figs. S2, S3). They share identical structures of their carboxyl ends, including the 5 cis-double bond position with most active endocannabinoids' AEA and 2-AG.
In view of the discovery of SEA and 2-SG in the CNS of the fads2 / mouse and their stoichiometric surrogate function of AEA and 2-AG, we further proved their ligand properties and activator function of the G protein-coupled CB1 receptor of the ECS in cb1-transfected HEK293 cells by monitoring the Fluo-4 fluorescence intensities evoked by Ca 2+ release. Ligand activity of SEA and 2-SG was comparable with that of AEA and 2-AG (Fig. 6).
Our findings expand the range of CB1 orthosteric ligand structures and underline their important molecular conformation essential for the binding and activation of the receptor. Our study provides valuable in vivo and in vitro systems for the translation of endocannabinoid structures designed and optimized in in silico analysis at the atomic level of modulators docking to crystallographic structures of the CB1 receptor (30)(31)(32).

Modification of the endocannabinoid pattern of brains of WT and fads2 / mice by sustained feeding experiments
The surprisingly stable residual LC-PUFA pool in the phospholipidome led to the synthesis of the LC-PUFA related endocannabinoids in nd-fads2 / brain (Fig. 3A). Sustained dietary supply of 6-AA suppresses the utilization of 3-PUFA in fads2 / mice and consequently led to reduced DHEA and 2-DG levels in phospholipids, in agreement with previous studies (33). Supplementation of 3-DHA in fads2 / mice suppressed the utilization of 6-PUFA and therefore reduces AEA and 2-AG levels. Reduction of AEA and 2-AG concentrations has also been observed in short-term dietary supplementation of 3-DHA and its impact on the ECS in brain and plasma in WT-CD1 mice reported previously (34).

Connectivity of the ECS and orexinergic system in
The nd-fads2 / mice (4 months) have a lean phenotype; 6-AAand 3-DHA-fads2 / mice develop pronounced obesity (18). The observed low levels of 2-AG in 3-DHA-WT and -fads2 / mice might contribute to the modulation of the orexinergic system, inducing enhanced lipogenesis in the DHA-fads2 / mice. HRP-and IHCstained coronal sections of nd-WT and -fads2 / brains display overlapping expression of CB1 and OX1R neurons (Fig. 5).
It is tempting to speculate that reduced availability of AEA and 2-AG in DHA-fads2 / mice suppresses CB1, TRPV1, and FAAH synthesis, documented in WB and IHC (Figs. 4,5).
In the brains of nd-fads2 / mice, the two novel endocannabinoids took over the function of AEA and 2-AG in the retrograde action as ligands of presynaptic CB1 on orexinergic neurons in different areas of brain (lateral hypothalamus, hippocampus, arcuate nucleus. The release of inhibitory GABAergic inputs promotes increased appetite and feeding and increased body fat mass, resulting in obesity, and excitatory glutamatergic inputs, suppressing hunger and induce hypophagy (14,35,36).
Our study delineates the individual role of EFAs and 3and 6-LC-PUFAs as essential dietary precursors in the synthesis of endocannabinoid ligands of CB1/2 receptors in the ECS, cooperating with the orexinergic neuronal network. However, beyond the regulation of the feeding behavior for maintaining energy homeostasis by the ECS and orexinergic system, additional relevant metabolic targets of PUFAs or their derivatives leading to the lean (anti-obese) phenotype of the LC-PUFA-deficient nd-fads2 / mice and the obese phenotype of the 3-DHA-fads2 / mice have to be addressed by future investigations. PUFAs as well as derived eicosanoids and endocannabinoids have been identified as CB1 and CB2 independent, directly binding ligands of subtypes of the nuclear receptor superfamily. 3-DHA and 6-AA have been recognized as ligands of nuclear receptor RxR in brain (51,52) and PPAR, -, and - (53,54).
This study predisposes the fads2 / mouse mutant as an unbiased mouse model for the discovery of therapeutically useful CB1 agonists and antagonists in the therapy of