ATGL is a biosynthetic enzyme for fatty acid esters of hydroxy fatty acids

Branched fatty acid (FA) esters of hydroxy FAs (HFAs; FAHFAs) are recently discovered lipids that are conserved from yeast to mammals1,2. A subfamily, palmitic acid esters of hydroxy stearic acids (PAHSAs), are anti-inflammatory and anti-diabetic1,3. Humans and mice with insulin resistance have lower PAHSA levels in subcutaneous adipose tissue and serum1. PAHSA administration improves glucose tolerance and insulin sensitivity and reduces inflammation in obesity, diabetes and immune-mediated diseases1,4–7. The enzyme(s) responsible for FAHFA biosynthesis in vivo remains unknown. Here we identified adipose triglyceride lipase (ATGL, also known as patatin-like phospholipase domain containing 2 (PNPLA2)) as a candidate biosynthetic enzyme for FAHFAs using chemical biology and proteomics. We discovered that recombinant ATGL uses a transacylation reaction that esterifies an HFA with a FA from triglyceride (TG) or diglyceride to produce FAHFAs. Overexpression of wild-type, but not catalytically dead, ATGL increases FAHFA biosynthesis. Chemical inhibition of ATGL or genetic deletion of Atgl inhibits FAHFA biosynthesis and reduces the levels of FAHFA and FAHFA-TG. Levels of endogenous and nascent FAHFAs and FAHFA-TGs are 80–90 per cent lower in adipose tissue of mice in which Atgl is knocked out specifically in the adipose tissue. Increasing TG levels by upregulating diacylglycerol acyltransferase (DGAT) activity promotes FAHFA biosynthesis, and decreasing DGAT activity inhibits it, reinforcing TGs as FAHFA precursors. ATGL biosynthetic transacylase activity is present in human adipose tissue underscoring its potential clinical relevance. In summary, we discovered the first, to our knowledge, biosynthetic enzyme that catalyses the formation of the FAHFA ester bond in mammals. Whereas ATGL lipase activity is well known, our data establish a paradigm shift demonstrating that ATGL transacylase activity is biologically important.

Branched fatty acid (FA) esters of hydroxy FAs (HFAs; FAHFAs) are recently discovered lipids that are conserved from yeast to mammals 1,2 . A subfamily, palmitic acid esters of hydroxy stearic acids (PAHSAs), are anti-inflammatory and anti-diabetic 1,3 . Humans and mice with insulin resistance have lower PAHSA levels in subcutaneous adipose tissue and serum 1 . PAHSA administration improves glucose tolerance and insulin sensitivity and reduces inflammation in obesity, diabetes and immune-mediated diseases 1,[4][5][6][7] . The enzyme(s) responsible for FAHFA biosynthesis in vivo remains unknown. Here we identified adipose triglyceride lipase (ATGL, also known as patatin-like phospholipase domain containing 2 (PNPLA2)) as a candidate biosynthetic enzyme for FAHFAs using chemical biology and proteomics. We discovered that recombinant ATGL uses a transacylation reaction that esterifies an HFA with a FA from triglyceride (TG) or diglyceride to produce FAHFAs. Overexpression of wild-type, but not catalytically dead, ATGL increases FAHFA biosynthesis. Chemical inhibition of ATGL or genetic deletion of Atgl inhibits FAHFA biosynthesis and reduces the levels of FAHFA and FAHFA-TG. Levels of endogenous and nascent FAHFAs and FAHFA-TGs are 80-90 per cent lower in adipose tissue of mice in which Atgl is knocked out specifically in the adipose tissue. Increasing TG levels by upregulating diacylglycerol acyltransferase (DGAT) activity promotes FAHFA biosynthesis, and decreasing DGAT activity inhibits it, reinforcing TGs as FAHFA precursors. ATGL biosynthetic transacylase activity is present in human adipose tissue underscoring its potential clinical relevance. In summary, we discovered the first, to our knowledge, biosynthetic enzyme that catalyses the formation of the FAHFA ester bond in mammals. Whereas ATGL lipase activity is well known, our data establish a paradigm shift demonstrating that ATGL transacylase activity is biologically important.
Branched FAHFAs are present in all mammalian tissues studied with the highest levels in white adipose tissue (WAT) and brown adipose tissue (BAT) 1 . Forty-six FAHFA families have been identified in mouse WAT 8 and even more in human WAT 3 , which have a unique composition of FAs and HFAs. In human WAT, these consist of 583 regioisomers 3 that are defined by the position of the ester bond between the HFA and the FA. FAHFAs are highly regulated in a tissue-specific manner in physiological and pathophysiological states. Subcutaneous (SQ) WAT FAHFA levels increase rapidly with fasting 1 . PAHSA levels in SQ WAT and serum correlate positively with insulin sensitivity in humans and mice, and levels are reduced before overt diabetes occurs in people with insulin resistance 1 . Specific PAHSA isomers are reduced in the breast milk of mothers who are obese 9 , and serum levels are decreased in patients with breast cancer 10 . Exercise training induces serum and WAT PAHSA levels in elderly women 11 . Levels of circulating palmitoleic acid ester of 9-HSA (9-POHSA) and oleic acid ester of 9-HSA (9-OAHSA) correlate with protective cardiovascular biomarkers in healthy humans 12 . Thus, FAHFAs have the potential to be important in many physiological and pathophysiological states in humans.
The biological effects of only a few FAHFAs are known. Administration of 5-and 9-PAHSA improves glucose tolerance and insulin sensitivity in mice by multiple mechanisms including enhancing insulin-mediated suppression of FA release from WAT, and improving insulin action to suppress hepatic glucose production 7 . PAHSA treatment also decreases WAT inflammation 1 and improves cognitive function in obese mice 13 . In addition, 5-and 9-PAHSA treatment reduces the incidence of auto-immune type 1 diabetes 6 and decreases the severity of colitis 4 in mice. PAHSAs and polyunsaturated FAHFA family members have anti-inflammatory effects directly on immune cells 1,14,15 . 9-Palmitic acid hydroxy palmitic acid and 9-oleic acid hydroxy palmitic acids seem Nature | Vol 606 | 30 June 2022 | 969 to increase basal metabolism and insulin sensitivity in mice 16,17 . Thus, FAHFAs are endogenous lipids with broadly beneficial metabolic and anti-inflammatory properties.
Little is known about the molecular mechanisms regulating FAHFAs. Whereas four hydrolytic enzymes have been identified-carboxyl ester lipase 18 , androgen-dependent tissue factor pathway inhibitor regulating protein, androgen-induced gene 1 (ref. 19 ) and hormone-sensitive lipase 20 -no enzymes that catalyse the formation of the ester bond between the FA and the HFA are known in an intact organism. Substrate availability seems to regulate FAHFA levels in vivo. For example, increased WAT glucose uptake and de novo lipogenesis markedly increases FAHFA levels in WAT and serum in mice overexpressing GLUT4 specifically in adipose tissue (AG4OX mice) 1 . Administration of C17-HFA to mice promotes FAHFA biosynthesis 1 , and availability of HFAs synthesized by cellular anti-oxidation pathways regulates FAHFA levels, as WAT PAHSA levels are reduced in peroxiredoxin 6 (Prdx6)-knockout (KO) mice 21 . Furthermore, supplementation of humans and mice with marine oils containing omega-3 polyunsaturated FAs increases FAHFAs consisting of these lipids, again demonstrating substrate regulation of FAHFA levels 14 . Other organisms can also synthesize FAHFAs from exogenous FAs. This is demonstrated by the fact that communal yeast strains that lack the FA synthesis gene, Fasn, and are dependent on exogenous FAs from human skin for survival, have endogenous FAHFAs 2 .
In this study, we aim to discover the enzyme(s) responsible for FAHFA biosynthesis (that is, the formation of the ester bond between the FA and the HFA). Understanding how FAHFAs are synthesized could lead to therapies to maintain or restore FAHFA levels in people with metabolic or immune-mediated disorders.

Higher FAHFA synthesis in AG4OX adipocytes
FAHFA levels are 8-16-fold higher in WAT of AG4OX mice 1 . We reasoned that this may result from upregulation of FAHFA biosynthesis. Basal levels of 9-PAHSA, 9-POHSA and 9-OAHSA are 1.5-3.4-fold higher in adipocytes derived from the stromal vascular fraction (SVF) of WAT of AG4OX mice (Fig. 1a-c). 9-hydroxy stearic acid (9-HSA) addition increased 9-FAHFA biosynthesis in a dose-dependent manner in both wild-type (WT) and AG4OX adipocytes (Fig. 1a-c). FAHFA biosynthesis was 35-50% higher in AG4OX compared to WT adipocytes at all HSA concentrations. Cis-10-heptadecenoic (C17:1) FA is a low-abundance endogenous FA that allows us to measure FAHFA production without a large background signal. It also ensures that differences in the biosynthetic activity are not due to different concentrations of FAs between AG4OX and WT cells. Addition of 9-HSA and C17:1 resulted in a twofold increase in 9-C17:1-HSA in AG4OX compared to WT cells (Fig. 1d)     Article biosynthetic activity, providing a cellular model to identify the specific biosynthetic enzymes and pathways.
Owing to MAFP specificity, the data indicate that an enzyme(s) with an active-site serine nucleophile mediates FAHFA biosynthesis. Several lipid-synthesizing enzyme classes have serine nucleophiles including acyltransferases and transacylases 22 .
To identify the enzyme(s) responsible for increased 9-FAHFA biosynthesis in AG4OX adipocytes, we performed activity-based protein profiling using a fluorophosphonate (FP)-alkyne probe 23,25 that uses the same functional group as MAFP and therefore should inhibit the same enzymes. Covalent labelling with the probe labels the enzyme with an alkyne that can be used to enrich and detect the target proteins through proteomics (Fig. 1g). Like MAFP, FP-alkyne blocked the increase in 9-PAHSA and 9-POHSA biosynthesis in AG4OX, but not WT, adipocytes (Extended Data Fig. 1c, d), validating that one or more FP-alkyne targets is involved in FAHFA biosynthesis.
We performed click chemistry with biotin azide to biotinylate the FP-alkyne target proteins in lysates from FP-alkyne-treated WT and AG4OX adipocytes and pulled down these proteins with streptavidin agarose 25 . The resulting samples were analysed by proteomics. We performed gene ontology analysis to filter the candidate proteins for enzymes involved in lipid metabolism, and then further truncated this list to include only proteins with an active-site serine nucleophile. These are listed in Extended Data Fig. 1e. ATGL had a robust proteomics signal and was increased by 1.5-fold in the AG4OX samples compared to WT, which we validated by western blotting (Fig. 1h and Extended Data Fig. 1f). ATGL, which catalyses the first step of TG hydrolysis to release FAs, is a critical regulator of lipid metabolism 26 .

ATGL overexpression increases FAHFA synthesis
To further demonstrate the importance of ATGL as a mediator of FAHFA biosynthesis, we overexpressed WT ATGL or the catalytically inactive serine mutant ATGL(S47A) in HEK293T cells and measured FAHFA biosynthesis. Both constructs were efficiently overexpressed (Extended Data Fig. 2a). Overexpression of WT ATGL increased FAHFA biosynthesis activity whereas enzymatically dead ATGL(S47A) did not (Fig. 2e). Thus, ATGL activity is sufficient to regulate FAHFA biosynthesis.

ATGL inhibition decreases FAHFA-TGs not glycerolipids
FAHFAs are incorporated into triglycerides (FAHFA-TGs) 27 . ATGL mediates FAHFA release from FAHFA-TGs during lipolysis 20,27,28 . To investigate whether loss of ATGL activity lowers FAHFA levels by trapping FAHFAs in FAHFA-TGs or alternatively by modulating substrate concentrations, we measured the levels of FAs, 9-HSA, FAHFAs, FAHFA-TGs and glycerolipids in the biosynthesis experiment. Atglistatin inhibited FAHFA biosynthesis by 76-81% (Fig. 2f). This was not due to accumulation of FAHFA-TGs, which were also diminished 67-78% (Fig. 2g). Thus, ATGL inhibition results in reduction of FAHFAs and FAHFA-TGs. Free FA (FFA) (Fig. 2h), triglyceride, diglyceride and monoglyceride levels (Extended Data Fig. 2b-d) were not changed in the presence of atglistatin, indicating that the marked reduction in FAHFA biosynthesis with ATGL inhibition is not a result of substrate depletion. These data demonstrate that ATGL regulates FAHFA biosynthesis through a biochemical mechanism that differs from the lipolytic effects of ATGL.

Atgl KO decreases FAHFA and FAHFA-TG synthesis
Biosynthesis of 9-FAHFAs (Fig. 2i, j), 12-FAHFAs (Extended Data Fig. 3a) and 5-FAHFAs (Extended Data Fig. 3b) is 60-84% reduced in adipocytes in which Atgl has been knocked out in the adipose tissue (AT-Atgl-KO adipocytes) versus Atgl fl/fl adipocytes (Fig. 2i, j), consistent with the atglistatin results. Nascent FAHFA-TGs are not detected in AT-Atgl-KO cells (Fig. 2k), suggesting that non-esterified FAHFAs are synthesized before they are esterified into TGs. Atglistatin has no impact on FAHFA biosynthesis in AT-Atgl-KO adipocytes ( Fig. 2i-k), confirming atglistatin specificity for ATGL inhibition. During the study C17:1 FA is incorporated into triglyceride as an acyl chain, but not D 20 -9-HSA (Fig. 2l). We cannot rule out that D 20 -9-HSA or 9-HSA might be incorporated into TGs under different conditions. Atgl fl/fl and AT-Atgl-KO adipocytes have similar levels of endogenous FFAs (PA, PO and HSA) (Fig. 2m). These data indicate that ATGL activity is necessary for FAHFA biosynthesis in adipocytes and loss of activity by small-molecule inhibition or genetic deletion reduces FAHFA biosynthesis. Furthermore, the reduction in FAHFA biosynthesis with ATGL inhibition or Atgl KO does not result from insufficient FFAs levels.

ATGL transacylation increases FAHFA synthesis
Having ruled out the possibility that FAHFAs are getting trapped in TGs, the data suggest that ATGL synthesizes FAHFAs through esterification of HSAs to FAs. In addition to its TG hydrolase activity 29-31 , ATGL has retinyl ester hydrolase activity 32 and phospholipase activity 33 , and produces triglycerides from monoglycerides and diglycerides through a CoA-independent transacylase reaction 30,34 in vitro. An ATGL transacylase activity could provide a mechanism for FAHFA biosynthesis, but the aforementioned reaction was observed only in vitro and has not been tested for FAHFA biosynthesis. ATGL is a member of the patatin-like phospholipase domain containing (PNPLA) family 26 . A family member, PNPLA1 has physiologically important transacylase activity that transfers an acyl chain from TG to a hydroxy-ceramide in the skin 35,36 .
To test whether ATGL can esterify HSAs, we incubated an affinity-purified truncation variant of ATGL including amino acids 1-288 (ATGL-288) (Extended Data Fig. 4a, b) and comparative gene identification-58 (CGI-58), an ATGL cofactor that increases its basal activity 37 , with 9-HSA and a triglyceride acyl donor, TG(18:1). ATGL-288 is a recently described form of ATGL that retains biological activity which is augmented by CGI-58 and inhibited by ATGLi 38 . We used ATGL-288 for all but one experiment because we can obtain large quantities of this owing to optimization of the purification and expression of this form of ATGL in bacteria 38 . CGI-58 had no FAHFA biosynthetic activity by itself (Fig. 3a, left). ATGL increased 9-OAHSA biosynthesis by 13-fold using TG as an acyl donor, and addition of CGI-58 further increased this transacylation activity by another 53% (Fig. 3a, left). Atglistatin inhibited this biosynthetic activity by 85%. This was a dose-dependent effect, underscoring the specificity of the activity for ATGL (Fig. 3a, right). The half-maximal inhibitory concentration (IC 50 ) of atglistatin for FAHFA biosynthesis catalysed by the full-length ATGL enzyme is 9.2 µM (Fig. 3a, right). ATGL catalysed robust 9-C17:1HSA biosynthesis in the presence of TG(17:1) but not C17:1 FFA (Fig. 3b).

Article
FAHFA biosynthesis is similar (Fig. 3c, d). The turnover rate of free FAHFA is about 420 pmol per mg protein per hour in our biosynthesis assay with ATGL-288 enzyme with acyl donor TG(18:1) (Fig. 3d). The turnover rate of free OA is about 52 nmol mg −1 protein per hour in our assay conditions (Fig. 3e).
These results show that the transacylase activity of ATGL synthesizes FAHFAs from di-and tri-glyceride acyl donors and HFAs. Furthermore, the same FAHFA biosynthesis pathway is active in human adipose tissue lysates (Fig. 3f), and human adipocytes (Fig. 3g) can synthesize FAHFAs from triglycerides and D 20 -9-HSA.

ATGL regulates FAHFA levels in vivo
To determine the physiological relevance of ATGL as a FAHFA biosynthetic enzyme in vivo, we studied mice lacking ATGL in adipose tissue (AT-Atgl-KO mice) 39 . Adipose-tissue-specific deletion of Atgl led to a marked decrease in 9-, 10-, 11-and 12/13-PAHSA regioisomers and a 78-94% decrease in total PAHSA levels in SQ WAT, perigonadal (PG) WAT and BAT (Fig. 3h-k) consistent with data in global-Atgl-KO mice 28 . By contrast, liver PAHSA levels are similar to those of control Atgl fl/fl mice (Fig. 3h, l). Serum 9-and 12/13-PAHSAs are lower and total serum PAHSA levels are decreased by 69% in AT-Atgl-KO mice (Fig. 3h, m). Levels of 5-PAHSA tended to decrease in PG WAT and serum in AT-Atgl-KO mice (Extended Data Fig. 5a, b). These results suggest that adipose tissue is a major contributor to circulating PAHSA levels. Similarly, OAHSA and POHSA/OAHPA levels are also decreased in AT-Atgl-KO SQ WAT, PG WAT and BAT by 92-98% but not in liver (Extended Data Fig. 5c-h). We also found that FAHFA-TG levels were 80-94% lower in WAT and BAT from AT-Atgl-KO mice, but unchanged in liver (Fig. 3n-q and Extended Data Fig. 5i-m). The different magnitude of reduction among specific FAHFA family members and regioisomers in adipose tissue of AT-Atgl-KO mice suggests that ATGL biosynthetic activity may be greater for certain regioisomers and family members. Triglyceride content per gram of tissue is increased in SQ WAT (by 43%) and BAT (by 172%) but not in PG WAT of AT-Atgl-KO mice compared with those of Atgl fl/fl mice (Fig. 3r). Consistent with the well-known function of ATGL as a lipase, SQ and PG WAT weight is increased in young AT-Atgl-KO mice (Extended Data Fig. 5n) as shown before 26,39 . Thus, the total amount of triglyceride content in the fat depots in AT-Atgl-KO mice is increased as expected in the absence of ATGL-mediated hydrolysis. The fact that an important substrate for FAHFA biosynthesis, triglycerides, is increased but levels of FAHFA and FAHFA-TGs are nearly undetectable strongly supports the conclusion that ATGL is a critical FAHFA biosynthetic enzyme in WAT and BAT. Triglyceride levels and tissue weight are decreased in AT-Atgl-KO liver ( Fig. 3r and Extended Data Fig. 5n), consistent with the literature 26,39 . Together, these data underscore the physiological importance of ATGL transacylase activity to regulate adipose FAHFA levels. This is a previously unknown function for this enzyme.
To provide direct evidence that ATGL catalyses FAHFA biosynthesis in vivo, we injected AT-Atgl-KO mice with D 20 -9-HSA (5 mg kg −1 , intraperitoneally) and measured the incorporation of this lipid into tissue FAHFAs as a direct measure of endogenous biosynthesis. This resulted in 90% lower 9-PA-D 20 HSA levels in PG WAT of AT-Atgl-KO compared to Atgl fl/fl mice (Fig. 3s). Percentage enrichment of 9-PA-D 20 HSA (that is, 9-PA-D 20 HSA divided by the sum of endogenous 9-PAHSA + 9-PA -D 20 HSA) is also markedly decreased, indicating that reduced FAHFA biosynthesis is responsible for lower endogenous FAHFA levels in PG WAT of AT-Atgl-KO mice (Fig. 3t). Newly synthesized 9-PA-D 20 HSA is not detectable in liver or serum of either genotype under these conditions (Fig. 3s). Nascent 9-PA-D 20 HSA-TGs were detected in WAT only in Atgl fl/fl mice (Fig. 3u). Levels of endogenous HSA substrates and enrichment of D 20 -9-HSA are similar in the WAT of Atgl fl/fl and AT-Atgl-KO mice (Fig. 3v).
The reduced FAHFA biosynthesis is cell autonomous, as it is also markedly reduced in AT-Atgl-KO adipocytes, resulting in 60-84% lower levels of nascent FAHFAs (Fig. 2i-k). These data prove that ATGL in adipose tissue is critical for FAHFA biosynthesis in vivo and lack of this activity results in lower endogenous adipose tissue and serum FAHFA levels in AT-Atgl-KO mice.
The lack of detection of newly synthesized FAHFAs and FAHFA-TGs in liver may be because they are below the detection limit for our liquid chromatography tandem mass spectrometry (LC-MS/MS) method. Endogenous FAHFA and FAHFA-TG levels in Atgl fl/fl mice are lower in the liver than in adipose tissue 19 (Fig. 3h and Extended Data Fig. 5g, h, l, m). Experiments with higher doses of D 20 -9-HSA (10-25 mg kg −1 ; Extended Data Fig. 6a) or longer duration (8 h) after injecting D 20 -9-HSA, revealed newly synthesized FAHFAs in the liver as well as the adipose tissue and serum. However, these FAHFAs could be synthesized in the liver or in other tissues and taken up by the liver. Human hepatocytes (HepG2 cells) can synthesize FAHFAs (Extended Data Fig. 6b), indicating that liver is capable of synthesizing FAHFAs.

DGAT drives ATGL-mediated FAHFA biosynthesis
In the current model, the addition of C17:1 FA would require C17:1 FA to be esterified into a triglyceride before transfer to HFA by ATGL. DGAT1 and DGAT2 are important for FA esterification into triglycerides 40 . DGAT1 inhibition resulted in a 30-71% decrease in FAHFA production (Fig. 4a-c), consistent with incorporation of FAs into TGs before they are transferred to HFAs to produce FAHFAs. Inhibition of both DGAT1 and ATGL or ATGL alone led to an 85-91% decrease in FAHFA biosynthesis (Fig. 4a-c). Furthermore, transfection of HEK293T cells with combinations of ATGL and DGAT1 demonstrates that these enzymes can work in tandem to drive FAHFA biosynthesis. Levels of nascent FAHFAs are Mean ± s.e.m. *P < 0.01, **P < 0.001, ***P < 0.0001 versus WT (one-way ANOVA). f, Model for ATGL transacylation activity synthesizing FAHFAs. Canonical ATGL lipase activity transfers an acyl chain of TG to a water molecule (H 2 O) thereby releasing a FA. In the presence of TG and HFA, ATGL transacylation activity transfers an acyl chain of TG to HFA, synthesizing FAHFA. Our data show that the lipase activity of ATGL (Fig. 3e) is much higher than its biosynthetic activity (Fig. 3d)

Discussion
We identified a biosynthetic enzyme for FAHFAs, ATGL, which we demonstrate catalyses the esterification of HFA and FA using TG, and to a lesser extent diglyceride, as the FA donor. We show that ATGL catalyzes the major FAHFA biosynthesis pathway in adipocytes and that ATGL inhibition decreases nascent FAHFAs and FAHFA-TGs without changing FA and glycerolipids levels. FAHFA and FAHFA-TG levels and biosynthesis are markedly reduced in WAT of AT-Atgl-KO mice. This results from loss of ATGL transacylation activity and not lipase activity, which would increase FAHFA-TGs. The observation that fasting and cold exposure increase FAHFA and FAHFA-TG levels in WAT of WT mice but not in global-Atgl-KO mice 1,28 further underscores that ATGL-catalysed FAHFA biosynthesis dominates over any potential hydrolytic activity of ATGL 20 or other lipases towards FAHFAs 20 or FAHFA-TGs 20 . Whereas ATGL can shuffle FAHFAs as acyl chains in TGs 20 , there is no evidence before our current data that ATGL can esterify HFAs with FAs (from TGs) to make FAHFAs. This is a paradigm shift in terms of the function of ATGL. Our observation that FAHFAs are synthesized before they are esterified into TGs is further evidence for the importance of ATGL transacylation activity. Mechanistically, both transacylase and lipase activities of ATGL require the formation of an acyl-enzyme intermediate, and this is followed by transfer of the acyl chain to a hydroxy group on an HFA (FAHFA biosynthesis) or to water (lipolysis; Fig. 4f).
Insulin-resistant states such as dietary obesity in mice and humans are associated with reduced adipose and/or serum PAHSA levels, and PAHSA administration to insulin-resistant mice enhances insulin sensitivity 1,5,7 . One might expect AT-Atgl-KO mice to be insulin resistant because PAHSA levels are low. However, adipose ATGL also has a crucial role in regulating FA availability to multiple tissues. For example, AT-Atgl-KO mice have reduced lipolysis that limits FA flux to the liver, resulting in decreased substrate availability for hepatic glucose production 39,40 . Similarly, FA availability to muscle and other tissues is limited, which can result in utilization of liver glycogen stores and hypoglycemia 26 . The fact that both adipose-specific-ATGL overexpression 41 and deletion improve insulin sensitivity in mice 26 reflects the complexity of the multiple metabolic effects of adipose ATGL.
Humans who are obese and resistant to insulin have lower ATGL levels in WAT compared to people who are obese and sensitive to insulin 42 , suggesting that ATGL may contribute to the regulation of FAHFAs in WAT and serum in people 1 . Thus, understanding the regulation of ATGL transacylase activity could lead to strategies to increase the levels of these beneficial lipids in disease conditions.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-04787-x.

Proteomics LC-MS and data analysis
The digested samples were analysed on an Orbitrap Fusion mass spectrometer (Thermo). Samples were injected directly onto a 25-cm, 100-µm ID column packed with BEH 1.7-µm C18 resin (Waters). Samples were separated at a flow rate of 300 nl min −1 on an nLC 1200 (Thermo). The LC solvents were as follows: buffer A, 0.1% formic acid in water; buffer B, 90% acetonitrile/0.1% formic acid. A typical LC run was 240 min long with a binary gradient consisting of the following steps: 1-25% buffer B over 180 min, 25-40% B over 40 min, 40-90% B over 10 min, 90% B for 10 min. The column was re-equilibrated with buffer A before the injection of sample. Peptides were eluted directly from the tip of the column and nanosprayed directly into the mass spectrometer by application of 2.5 kV voltage at the back of the column. The mass spectrometer was operated in a data-dependent mode. Full MS1 scans were collected in the Orbitrap at 120,000 resolution. The cycle time was set to 3 s, and within this 3 s the most abundant ions per scan were selected for collision-induced dissociation MS/MS in the ion trap. Monoisotopic precursor selection was enabled and dynamic exclusion was used with an exclusion duration of 5 s.
Protein and peptide identification were performed with Integrated Proteomics Pipeline (IP2, Integrated Proteomics Applications). Tandem mass spectra were extracted from raw files using RawConverter 51 and searched with ProLuCID 52 against the UniProt mouse database. The search space included all fully tryptic and half-tryptic peptide candidates. Carbamidomethylation on cysteine was considered as a static modification. Data were searched with a 50-ppm precursor ion tolerance and a 600-ppm fragment ion tolerance. Identified proteins were filtered to a 10-ppm precursor ion tolerance using DTASelect 53 utilizing a target-decoy database search strategy to control the false discovery rate to 1% at the protein level 54 .

Western blotting
Frozen adipose tissue (human or mouse) and HEK293T cells were dounce homogenized on ice in buffer A (0.25 M sucrose, 1 mM EDTA, 1 mM dithiothreitol, 20 µg ml −1 leupeptin, 2 µg ml −1 antipain and 1 µg ml −1 pepstatin, pH 7.0). Lysates were passed through a 26-G or 30-G needle attached to a 1-cc syringe 30 times to break open the cells. Samples were sonicated in a 550 Sonic Dismembrator Cup Horn (Fisher Scientific) filled with ice-cold water for 5 min. Homogenates were centrifuged to remove nuclei and unbroken cells (1,000g, 2 °C, 15 min). Protein concentrations were determined using the reducing reagent bicinchoninic acid assay (Thermo Fisher). Fifty micrograms of proteins were separated by SDS-polyacrylamide gel electrophoresis (7.5% Bio-Rad TGX gels) and transferred to nitrocellulose membranes. Membranes were stained using Revert 700 (Licor) to visualize and quantify total proteins on the blots. Blots were blocked with 5% milk in TBS for 30 min. Membranes were probed with antibodies to ATGL (Cell Signaling number 2439, 30A4, 1:2,000), DGAT1 and DGAT2 (gift from R. V. Farese's laboratory, 1:1,000) 47 and GAPDH (Cell Signaling number 2118, 14C10, 1:5,000). IR Dye 800CW anti-rabbit and anti-mouse secondary antibodies (LI-COR,1:15,000) were used as secondary antibodies for visualization. The uncropped and unprocessed scans are shown in Supplementary Figs. 1-3. Band intensities were quantified using Image Studio Lite software.
Using 99% pure TG(18:1) substrate from Sigma, compared to 95% pure TG(18:1), from Cayman resulted in a higher rate of 9-OAHSA synthesis in our in vitro biosynthesis experiments. As all comparisons were made within the same experiment, this did not affect the relative values within an experiment. For the IC 50 calculation in Fig. 3a (right), we used a nonlinear regression log[inhibitor] versus response (four parameters) model and used the mean value of the biosynthesis assay with no enzyme as the bottom constraint.

Human SQ WAT lysate FAHFA biosynthesis assay
The assay was performed as described above, except 400 µg quantities of human SQ WAT lysates from three different donors were incubated with substrates TG and D 20 -9-HSA for 60 min at 37 °C while shaking.

Mouse studies
All experimental procedures were approved by the Institutional Animal Care and Use Committee of Beth Israel Deaconess Medical Center and performed following its policies. Mice were housed at the Beth Israel Deaconess Medical Center at 23.3 °C temperature and 40-60% humidity, on ventilated racks (25 ACH) under a 12 h light/12 h dark cycle and fed on chow diet (Lab Diet, 5008). AG4OX mice were generated in our laboratory as previously described 39,56 . AG4OX and control WT female and male mice (FVB background) of 9-10 weeks of age were used for SVF adipocyte preparations. AT-Atgl fl/fl (C57BL/6J) mice were purchased from Jackson Laboratory (stock number 024278) and crossed to adiponectin-Cre mice (C57BL/6J), a gift from E. D. Rosen 57 (Beth Israel Deaconess Medical Center) to generate AT-Atgl-KO mice. The metabolic phenotype of AT-Atgl-KO mice was previously described 39 and was confirmed in our colony. AT-Atgl-KO and littermate Atgl fl/fl control female mice of 10-11 weeks of age were used for studies of endogenous FAHFA levels in tissues and in vivo measurement of FAHFA biosynthesis. Animals were euthanized by CO 2 inhalation, blood was collected by cardiac puncture to obtain serum, and tissues were rapidly dissected, flash-frozen in liquid nitrogen and stored at −80 °C.

FAHFA biosynthesis in vivo
For in vivo experiments, mice were randomly assigned to different groups. The experiment was performed between 08:00 and 13:15. Ad-libitum-fed AT-Atgl-KO and Atgl fl/fl control mice were intraperitoneally injected with 5 mg kg −1 D 20 -9-HSA in 49.5% H 2 O/ 0.5% Tween 20, PEG 400/50% vehicle. The volume of the injection was adjusted on the basis of body weight (5 µl g −1 body weight; that is, 100 µl for a 20-g mouse). Food was immediately removed. In our experience with this strain of mice, 5-h food removal in the morning does not increase serum non-esterified FA levels (ad libitum fed 0.68 ± 0.03; after 5-h food removal 0.75 ± 0.04 mmol l −1 , n = 7 per group), which is consistent with no significant increase in lipolysis. Injections were staggered by 7 min, so each mouse was euthanized exactly 4 h after substrate injection for tissue and serum collection. Serum, liver and PG WAT lipids were extracted and D 20 -9-HSA incorporation into FAHFAs and TG-esterified FAHFAs was measured by LC-MS. A further biosynthesis study with a D 20 -9-HSA dose of 25 mg kg −1 (Extended Data Fig. 6a) was also performed in the same manner in WT mice.

Lipid extraction
For the measurement of FAHFAs from cultured cells, total lipids were extracted using the modified Bligh-Dyer method 58 . In brief, 1.5 ml of 2:1 chloroform/methanol with the internal standard [ 13 C 16 ]9-PAHSA (5 pmol) was added to 500 µl of cell suspension in PBS. Samples were vortexed and centrifuged at 2,000g for 7 min. The bottom organic phase was transferred into a new vial, and dried under a N 2 stream.
For tissues (50-75 mg) and serum (100 µl), total lipids were extracted as described above. FAHFAs were enriched using a solid-phase extraction (SPE) column (Hypersep silica 500 mg) as described previously 59 . Briefly, columns were equilibrated with 15 ml hexane, and the samples were then resuspended in 200 µl chloroform and loaded on the SPE column. The neutral lipid fraction containing TGs was eluted with 16 ml of 95:5% hexane/ethyl acetate, and the polar lipid fraction containing non-esterified FAHFAs was eluted using 15 ml ethyl acetate. The FAHFA fractions were dried under a stream of N 2 and stored at −40 °C until LC-MS/MS analysis. Total TG levels were quantified from the same piece of tissue as FAHFAs. Neutral lipid fractions of SPE eluents were used to quantify total tissue TG levels using the Infinity triglyceride colorimetric assay kit (Thermo, TR22421) as previously described 60 .
TG-esterified FAHFAs were extracted, hydrolysed and enriched by multi-step lipid extractions and SPE fractionations with a mild LiOH hydrolysis method as described elsewhere 27 . Internal standard TG ([ 13 C 16 ]PAHSA/16:0/16:0) was added at the beginning of sample processing and D 31 -PAHSA internal standard was added after 24-h mild hydrolysis of neutral fractions. The amount of internal standard was adjusted on the basis of the type of tissue or cell. Samples were stored at −40 °C until FAHFAs released by mild hydrolysis (TG-esterified FAHFAs) were measured by LC-MS analysis.

Sample analysis using LC-MS/MS
FAHFA isomers were quantified using an Agilent 6470 Triple Quad LC-MS/MS instrument through multiple reaction monitoring (MRM) in the negative ionization mode as described previously 27 . In brief, cell culture lipid extracts were reconstituted in 50 µl methanol, and tissue extracts were reconstituted in 100 µl methanol. A 7 µl volume of sample was injected onto a UPLC BEH C18 Column (Waters Acquity, 186002352). FAHFA regioisomers from different FAHFA families were resolved using a 93:7 methanol/water with 5 mM ammonium acetate and 0.01% ammonium hydroxide solvent through an isocratic gradient at 0.15 ml min −1 flow rate for 45 min. Transitions for targeted FAHFAs are listed in Supplementary Table 1. MS acquisition parameters for tandem MS were: gas temperature = 250 °C, gas flow = 12 l min −1 , nebulizer = 20 psi, sheath gas temperature = 250 °C, sheath gas flow = 11 l min −1 . Spray voltage was −1.0 kV.

Quantification of FA/HSA and D 20 -9-HSA for in vivo/intact cells and in vitro biosynthesis studies
A 25 mg quantity of PG WAT was homogenized in 1.5:1.5:3 ml PBS/methanol/chloroform with the internal standard 20 pmol 13 C 9 -9-HSA. Samples were vortexed and centrifuged at 2,000g for 7 min. Bottom organic phase was transferred into a new vial, dried under a stream of N 2 .

Semi-targeted lipidomics in 3T3-L1 adipocyte lipid extraction and LC-MS
Lipids were extracted using a modified version of the Bligh-Dyer method 58 . In brief, samples were manually shaken in a glass vial (VWR) with 1 ml PBS, 1 ml methanol and 2 ml chloroform containing internal standards ([ 13 C 16 ]palmitic acid and D7-cholesterol) for 30 s. The resulting mixture was vortexed for 15 s and centrifuged at 2,400g for 6 min to induce phase separation. The organic (bottom) layer was retrieved using a Pasteur pipette, dried under a gentle stream of nitrogen, and reconstituted in 2:1 chloroform/methanol for LC-MS analysis.
Lipidomics analysis was performed on a Vanquish HPLC online with a Q-Exactive quadrupole-Orbitrap mass spectrometer equipped with an electrospray ion source (Thermo). Data were acquired in positive and negative ionization modes. Solvent A consisted of 95:5 water/methanol; solvent B was 70:25:5 isopropanol/methanol/water. For positive mode, solvents A and B contained 5 mM ammonium formate with 0.1% formic acid; for negative mode, solvents contained 0.028% ammonium hydroxide. A Bridge (Waters) C8 column (5 µm, 4.6 mm × 50 mm) was used. The gradient was held at 0% B between 0 and 5 min, increased to 20% B at 5.1 min, increased linearly from 20% to 100% B between 5.1 and 55 min, held at 100% B between 55 min and 63 min, returned to 0% B at 63.1 min, and held at 0% B until 70 min. The flow rate was 0.1 ml min −1 from 0 to 5 min, 0.3 ml min −1 between 5.1 min and 55 min, and 0.4 ml min −1 between 55 min and 70 min. The spray voltage was 3.5 kV and 2.5 kV for the positive and negative ionization modes, respectively; the S-lens RF level was 65. The sheath, auxiliary and sweep gases were 50, 10 and 1, respectively. The capillary temperature was 325 °C and the auxiliary gas heater temperature was 200 °C. Data were collected in full MS/dd-MS2 (top 10) mode. The full MS scan was acquired in the range 150-1,500 m/z with a resolution of 70,000, an automatic gain control (AGC) target of 1 × 10 6 and a maximum injection time of 100 ms. MS2 was acquired with a resolution of 17,500, a fixed first mass of 50 m/z, an AGC target of 1 × 10 5 and a maximum injection time of 200 ms. Stepped normalized collision energies were 20, 30 and 40%. The inclusion list was on, and the instrument was set to pick other ions when idle.
A lipid target list was generated with LipidCreator. Mass accuracy, chromatography retention time and peak integration of all targeted lipids were verified with Skyline 61 . Peak areas were used in data reporting, and data were normalized using internal standards.

Statistical analyses
Data are shown as individual data points and mean ± s.e.m. The numbers of participants (human WAT lysate study), mice and wells of cells per group are annotated as n. Each cell culture biosynthesis experiment was replicated with similar results at least twice. Graphs and statistical analyses were generated using GraphPad Prism 6.0 and 8.0 (GraphPad Software). Specific statistical tests and n used for each experiment are listed in the figure legends. Data were analysed using unpaired t-tests, two-tailed for single comparisons, t-test corrected for Holm-Sidak multiple comparison, one-way ANOVA for multiple comparisons (Holm-Sidak multiple comparison) and two-way ANOVAs for comparison of genotype and treatments (Holm-Sidak multiple comparison) as listed in the figure legends.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
The datasets and materials generated during the current study are available from the corresponding author on reasonable request. Source data are provided with this paper.