Study was performed in accordance with the protocols approved by the ethical committee of the Medical University of Graz (Vote no: 29-319 ex 16/17 and 24-529 ex 11/12). All subjects gave written informed consent. Placentas from caesarean section and vaginal delivery were used within 20 min after delivery. Important subject characteristics of this study cohort are depicted in Table 1 . In order to collect tissue samples, the placenta was divided into quadrants and a cross sectioned piece of 7-10 mm diameter was scissored from each quadrant. Tissue samples were either snap frozen in liquid nitrogen and stored at −80°C for protein isolation, or formalin fixed and embedded into paraffin for immunohistochemistry.

Frozen 20-30 mg placental tissue pieces were homogenized in 700μl Qiazol lysis reagent (Qiagen, Cat# 217004) for 20 seconds 6500 revolutions per minute, by MagnaLyser (Roche, Basel, Switzerland) followed by 1 minute on ice and repeated 3 times. Next, total RNA content from cells and tissue lysates were isolated using the RNeasy^®Mini Kit (Qiagen, Cat# 217004). Reverse transcription was performed using 1μg of RNA and LUNA Script RT SuperMix Kit (New England Biolabs, Cat#E3010L). For RT-qPCR analysis LUNA Universal qPCR Master Mix (New England Biolabs, Cat#M3003E) and BioRad CFX384 Touch Syllabus were used. QuantiTect Primer Assays were used for gene amplification. For 18S reference gene amplification custom DNA oligos were designed (F-(5 ‘-3 ‘) CTACCACATCCAAGGAAGCA/R-(5 ‘-3 ‘) TTTTTCGTCACTACCTCCCCG). The expression of target genes DAGLα (GeneGlobe ID - QT00038164) and DAGLβ (GeneGlobe ID - QT00074319) was normalized to reference genes 18S, RPL30 (GeneGlobe ID - QT00056651) and HPRT1 (GeneGlobe ID - QT00059066). Target and reference gene ΔCT values are corrected for respective primer efficacy.

To detect and discriminate DAGLα and DAGLβ mRNA on cellular level, RNAscope^® 2.5 HD Reagent Kit-RED assay (Advanced Cell Diagnostics, Cat#PN 322350) was used according to the manufacturer’s protocol. In short, 5 µm thick formaldehyde-fixed paraffin-embedded sections were de-paraffinized and pre-treated under standard pre-treatment conditions with hydrogen peroxide, target retrieval reagents and protease solution. The sections were covered with probe solution and incubated for 2 hours at 40°C using the HybEZ Hybridization System (Advanced Cell Diagnostics, Cat#PN 321710/321720). The sections were treated with AMP 1 to 6 according to the manufacturer´s manual, using the HybEZ Hybridization System. The multi-step hybridization process included hybridization to alkaline phosphatase-labeled probes and resulted in the detection of signal using Fast Red as a substrate. To combine ISH with immunohistochemistry (IHC), after performing ISH IHC was started from the blocking step as described below.

Placental tissue sections were blocked with 4% BSA and 10% secondary antibody host serum in PBS/0.3% Triton X100 and incubated overnight with primary antibody solutions. Primary antibodies for cytokeratin 7 (1:500, Abgent, Cat#AJ1229a), CD163 (1:200, Thermo Fisher Scientific, Cat#MA1-82342), and Von Willebrand Factor (1:500, Dako, Cat#A0082) were used. To detect Cytokeratin 7 (CK7) and Von Willebrand Factor (VWF) goat anti-rabbit Alexa Fluor 647 secondary antibody was used (1:500, Cell Signaling Technology, Cat#4414) and displayed in white. CD163 primary antibody incubation was followed by goat anti- mouse Alexa Fluor 488 secondary antibody application (1:500, Invitrogen, Cat#A32723) and displayed in green. Sections were counterstained with DAPI, sealed with a coverslip using VECTASHIELD^® Antifade Mounting Medium with DAPI (Vector Laboratories, Cat# H-1200-10) and stored at 4°C until imaged. Representative images were captured on Nikon A1 confocal microscope (original magnification ×40) and prepared using FIJI software v.1.51h.

For quantitative determination and localization analysis, ten z-stacks of each section were acquired using a Nikon A1 confocal with a ×40 objective at a step size of 0.5 µm. An automated image analysis was created with the software package FIJI v.1.51h. The analysis entailed a basic pre-processing, generating a maximum intensity projection and mean filter smoothing, followed by application of an algorithm-based threshold. Feature detection of channels containing CK7 or ISH information was achieved employing RenyiEntropy (25), IsoData for the CD163 channel (26) and Otsu for the VWF containing channel (27). Generated regions of interest (ROI) of the ISH were separated by watershed and counted the ROIs of the remaining channels were used to determine the cell type specific ISH localization due to overlap.

Gel-based ABPP experiments were performed as previously described (28). Frozen tissues were thawed on ice and homogenized in cold lysis buffer (20 mM HEPES pH 7.2, 250 mM sucrose, 1 mM MgCl2, 2.5 U/mL benzonase). After incubation on ice for 15 min, tissue debris was pelleted by centrifugation (2500 × g, 3 min, 4°C) and supernatant was transferred to a clean tube. Subsequently the supernatant was centrifuged at 30.000 × g (90-120 min, 4°C) to pellet the membrane-associated fraction and separate it from the soluble proteome. After removal of the soluble supernatant, the membrane pellet was washed with cold HEPES buffer (20 mM, pH 7.2) followed by resuspension in cold HEPES buffer by pipetting. Concentration of membrane-associated and soluble proteome was quantified (Bradford; BioRad Technologies, CA, USA) and adjusted to desired concentration (2 mg/mL) in HEPES buffer (20 mM, pH 7.2). For direct labeling, proteomes were sequentially treated with one-step activity-based probes DH379 (30 min, 1 µM, Cy3, RT) and FP-Bodipy (15 min, 500 nM, Cy2, RT) or MB064 (30 min, 250 nM, Cy3, RT) alone in a 15 µL total reaction volume. For competitive ABPP experiments, this step was preceded by incubation with DH376 in vitro or ex vivo and LEI-105 in vitro at indicated concentrations. The reactions were quenched by the addition of 5 µL 4x Laemmli-buffer (BioRad Technologies, CA, USA). After separation by SDS-PAGE (10% acrylamide) at 180V for 75 min, samples were visualized by in-gel fluorescence scanning (Cy2 532/28, Cy3 605/50, Cy5 700/50 filter settings) using a flatbed fluorescent scanner ChemiDoc™ MP Imaging System (Bio-Rad, Hercules, CA, USA). Coomassie staining was used to control the protein loading. Gel fluorescence is shown in greyscale, and optical density of the signals was determined using ImageLab 6.1 Biorad.

Membrane fractions of placental tissues were prepared as described above (see Gel-based activity-based protein profiling (ABPP)). Membrane lysates were diluted to 10 ng/µL in assay buffer (50 mM HEPES pH 7.5, 0.0025% Triton X-100). Fluorescent measurements were carried out at RT in a black flat bottom 96-well plate (Thermo Fisher Scientific, MA, USA) in the presence of 0.5 µM EnzChek™ lipase substrate (Thermo Fisher Scientific, Cat#E33955) in 100 μL final volume using a Clariostar plate reader (BMG Labtech, Germany) and excitation/emission wavelengths of 477/525. For competitive experiments placental membrane lysates were pre-incubated with DH376 (100 nM) and LEI-105 (1 µM) for 30 min at RT, respectively. DMSO served as vehicle control and denatured samples (1% SDS, 5 min, 100°C) served as background controls. Background substrate hydrolysis was deducted from each measurement. Each data point is the mean of three technical replicates of n=3 placentas, for concentration testing and n=4 placentas for competitive experiments. The slope t=10 min to t=60 min was used as the enzymatic rate (RFU/min). Enzyme kinetics were plotted as curves in Graph Pad Prism 9 Software (GraphPad Software Inc., CA, USA).

Placental tissues were homogenized and prepared as described in Gel-based activity-based protein profiling (ABPP) section. The chemical proteomics workflow is based on previously published protocol (29) and conducted with minor modifications. In short, cytosolic and membrane fractions of placental tissue lysates (250 µg protein, 1 mg/mL, n=5) were incubated with serine hydrolase probe cocktail (10 µM MB108, 10 µM FP-Biotin, 30 min, 37 ˚C, 300 rpm). A pool of denatured vehicle control samples (1% SDS, 5 min, 100°C) was taken along as a negative control. Following steps were preformed according to protocol, including precipitation, alkylation, avidin enrichment, on-bead digestion, and sample preparation. Dried and desalted peptide samples were stored at -20°C until LC-MS analysis. Prior to measurement, samples were reconstituted in 50 µL 97:3:0.1 solution (H2O, ACN, FA) containing 10 fmol/µL yeast enolase digest (Waters, cat# 186002325) and transferred to LC-MS vials. Additionally, a quality control sample was prepared to prevent overloading the nanoLC system and the automatic gain control (AGC) of the QExactive HF mass spectrometer. LC-MS data was analyzed by MaxQuant software 2.0 applying match between runs. For further analysis, the following cut-offs were used: unique peptides ≥ 2, identified peptides ≥ 2, ratio positive over negative control ≥ 2. Additionally, targets were filtered against a putative probe-target list including human metabolic serine hydrolases.

Placental perfusion setup is based on the setup published by Schneider et al. (30) and adapted as published by Hirschmugl et al. (31). In short, within 30 min after delivery of the placenta a single placental cotyledon chorionic-artery and vein pair was cannulated and flushed with perfusion buffer, containing DMEM (DMEM, phenol red free, Gibco by Life Technologies, ThermoFisher Scientific, MA, USA) mixed (3:1) with Earl's buffer (6.8 g/L NaCl, 0.4 g/L KCl, 0.14 g/L NaH2PO4, 0.2 g/L MgSO4•7H2O, 0.2 g/L CaCl2, 2.2 g/L NaHCO3, pH 7.4, all Merck, Darmstadt, Germany), amoxicillin (250 mg/L, Sigma-Aldrich, Steinheim, Germany), glucose (2 g/L Merck, Darmstadt, Germany), and essential fatty acid free bovine serum albumin (BSA) (35 g/L, Sigma-Aldrich, Seinheim, Germany). The cannulated cotyledon and surrounding tissue were placed in the pre-warmed perfusion chamber and the fetal circulation was connected to a magnetic pump (Codan, Salzburg, Austria) with a constant fetal artery inflow of 3 mL/min. The perfusion buffer is constantly fumigated by a gas exchange device (LivingSystems, St. Albans, VT, US) operated with 95% N2 and 5% CO2 on the fetal site during the experiment. A micro catheter pressure sensor (Millar, US) inserted into the fetal arterial cannula recorded the backflow pressure which should not exceed an average of 65 mbar. The impermeability of the perfused cotyledon was monitored within the first 30 min and each cotyledon displayed at least 95% fetal flow recovery. The maternal circulation was established by inserting three rounded needles into the intervillous space of the cotyledon with a flow rate of 9 mL/min. During the experiment, the perfusion buffer was gassed with 5% CO₂, 20% O₂ and 75% N₂ through the gas exchanging device. DH376 (1 µM) was added to the fetal and maternal perfusion buffer reservoirs and the system changed to closed circuit in all inhibitor experiments. During the experiment maternal and fetal perfusates were collected every 30 min via a sampling port and oxygen (pO₂), carbon dioxide (pCO₂), pH, lactate production, and glucose consumption measurements were applied by a blood gas analyzer (Radiometer, Copenhagen, Denmark). The data sets obtained by the blood gas analyzer, magnetic pumps, and pressure sensor were registered and recorded via LabVIEWbased recording software (Beko engineering, Graz, Austria). After 4h of closed perfusion time, samples of both circuits were collected, centrifuged, and stored at −80°C. The perfused placental tissue was processed in cold PBS and snap frozen in liquid nitrogen until further analysis.

Placenta (pl, ~10 mg powdered), and perfusate (pf, 140 µl) samples were extracted according to Matyash et al. (32). In brief, samples were homogenized using two beads (stainless steel, 6 mm) on a Mixer Mill (Retsch, Haan, GER; 2x10sec, frequency 30/s) in 700 µl methyl-tert-butyl ether (MTBE)/methanol (3/1, v/v) containing 500 pmol butylated hydroxytoluene, 1% acetic acid, and internal standards (IS; pl: 20 pmol 15:0/15:0/15:0 triacylglycerol, 13 pmol rac-17:0/17:0 diacylglycerol, rac-17:0 monoacylglycerol, 50 pmol 17:0/17:0 phosphatidylcholine, Larodan, Solna, Sweden; 133 pmol 17:0/17:0 phosphatidylethanolamine, 30 pmol 17:0/17:0 phosphatidylserine, 8 pmol 17:1 lyso-phosphatidylcholine, 30 pmol 17:1 lyso-phosphatidylethanolamine, Avanti Polar Lipids, Alabaster, AL, USA; cb and pf: 2 nmol 17:0 FA, 800 pmol C21:0 FA, Sigma-Aldrich, St. Louis, MO, USA). Total lipid extraction was performed under constant shaking for 30 min at RT. After addition of 140 µl dH₂O (pl) and further incubation for 30 min at RT, samples were centrifuged at 1,000 x g for 15 min. 500 µl of the upper, organic phase were collected and dried under a stream of nitrogen. Lipids were resolved in 500 µl MTBE/methanol (3/1, v/v). Pl extracts were diluted 1:4 in 2-propanol/methanol/dH₂O (SolA; 7/2.5/1, v/v/v) for LC-MS analysis. To determine fatty acid levels, 200 µl (pf) were derivatized according to Bollinger et al. (33) using the AMP+ MS Kit (Cayman Chemical, Michigan, USA) and resolved in 500 µl SolA for LC-MS analysis. Protein precipitates of the extractions were dried, solubilized in NaOH (0.3 N) at 65°C for 4 h and the protein content was determined using Pierce™ BCA reagent (Thermo Fisher Scientific, MA, USA) and BSA as standard. Chromatographic separation was performed on a 1290 Infinity II LC system (Agilent, CA, USA) equipped with Zorbax RRHD Extend-C18 column (2.1x50 mm, 1.8 µm; Agilent, CA, USA) running a 10 min linear gradient from 60% solvent A (H2O; 10 mM ammonium acetate, 0.1% formic acid, 8 µM phosphoric acid) to 100% solvent B (2-propanol; 10 mM ammonium acetate, 0.1% formic acid, 8 µM phosphoric acid). The column compartment was kept on 50°C. A 6470 Triple Quadrupole mass spectrometer (Agilent, CA, USA) equipped with an ESI source was used for detection of lipids in positive mode. Data acquisition was done by MassHunter Data Acquisition software (B.10, Agilent, CA, USA) either in MRM (glycerol- and glycerophospholipids) or SIM (fatty acid derivatives) mode. Lipidomic data were processed using MassHunter Workstation Quantitative Analysis for QQQ (V.9, Agilent, CA, USA), normalized for recovery, extraction-, and ionization efficacy by calculating analyte/IS ratios (AU) and expressed as AU/µg protein.

Graph Pad Prism 9.02 Software (GraphPad Software Inc., CA, USA) was used for statistical analysis and graph plotting. Data are presented as mean ± SEM. All obtained datasets were tested for normal distribution with the Shapiro-Wilk and Kolmogorov-Smirnov test. Depending on the distribution of datasets parametric or non- parametric statistical tests were applied. If two or more normal distributed groups were compared student`s t-test or one-way ANOVA, including Benjamini- Hochberg post-hoc was performed. If the dataset did not show a normal distribution Mann-Whitney U test or Kruskal-Wallis test followed by Benjamini- Hochberg post hoc was applied. Two-way ANOVA was applied comparing two or more groups including different variables using Benjamini- Hochberg post hoc for multiple comparison correction. All herein presented p-values correspond to a FDR of 1% for multiple testing and p-values below 0.05 were considered statistically significant.

To elaborate DAGL expression in placental tissues, we first determined DAGLα/β mRNA levels by RT-qPCR. In comparison to DAGLα substantially higher DAGLβ expression was observed ( Figure 1A , p<0.0001). Serine hydrolase activities in the placenta were examined by gel-based ABPP. To obtain a broad view of placental serine hydrolase activities and DAGL-specific signals, we used a probe cocktail of non- selective FP-Bodipy (34) and the DAGLα/β directed fluorescent probe DH379 (28). Using FP-Bodipy, we detected a broad spectrum of active enzymes in placental lysates. Application of DH379 visualized DAGLβ at the expected molecular weight of ~70 kDa ( Figures 1B, C ). To confirm the presence of active DAGLβ, competitive ABPP was applied using the DAGL inhibitors DH376 (IC₅₀ 3–8 nM) (28) and LEI-105 (IC₅₀ ~32 nM) (35). Application of DH376 and LEI-105 reduced DAGLβ activity in a dose-dependent manner. DH376 led to full inhibition at the lowest concentration of 0.1 µM ( Figure 1B ), while LEI-105 treatment led to a substantial reduction of DAGLβ signals at a concentration of 0.5 µM ( Figure 1C ). Notably, we were not able to detect DAGLα activity at the expected molecular weight of ~120 kDa, using the DAGLα tailored probe MB064 (36), likely due to low expression levels compared to DAGLβ ( Supplementary Figure 1 ). DAGLβ activity in placenta tissue lysates was also investigated using the commercially available lipase substrate EnzChek™ ( Figure 1D ). We applied 100 nM of DH376, which represented the lowest inhibitor concentration leading to potent enzyme inhibition of DAGLβ in-gel ( Figure 1B ). In comparison to DH376, LEI-105 exhibits lower activity against DAG-lipases and acts as a reversible enzyme inhibitor. Although a substantial reduction in DAGLβ enzyme activity was observed by application of 0.5 µM ( Figure 1C ), we applied 1 µM LEI-105 to ensure complete enzyme inhibition over time. The application of LEI-105 and DH376 reduced hydrolase activity by 51% and 70%, respectively ( Figure 1E ). The differences in inhibitor efficacy can be explained by previous observations showing that LEI-105 exhibits higher selectivity for DAGLβ than DH376. At the used inhibitor concentration, DH376 is expected to inhibit α/β hydrolase domain-containing protein 6 (ABHD6), carboxylesterase 1 and 2 (CES1/2), and hormone- sensitive lipase (HSL) (28), which can contribute to the hydrolysis of the EnzChek™ substrate. Gel-based ABPP experiments enabled us to detect specific DAGLβ activity in placental tissue and we could further decipher DAGL-dependent substrate hydrolysis by administration of DH376 and LEI-105. Overall, our observations indicate that the placenta predominantly expresses DAGLβ, which consequently can affect DAG, MAG, and fatty acid (FA) metabolism.

ABPP using FP-Bodipy already suggested that the placenta expresses a broad spectrum of serine hydrolases ( Figures 1B, C ). To get a profile of these enzymes, we performed mass spectrometry-based chemical proteomics utilizing the biotinylated non- selective probes MB108 and FP-Biotin for target identification. While gel-based ABPP experiments strongly rely on specific inhibitors for target identification, MS- based ABPP enables target enrichment and provides high sensitivity. To increase the resolution of proteins, tissue lysates were separated into membrane and cytosolic fraction. This approach resulted in the identification of 38 and 33 different serine hydrolases in membrane and cytosolic fractions, respectively ( Figure 2 ). Activities of several α/β hydrolase domain-containing protein family members (ABHD) and phospholipases such as DDHD2, patatin-like phospholipase domain-containing proteins (PNPLA) and members of the phospholipase A2 family (PLA2) were identified. Furthermore, lipases involved in DAG, MAG and FA metabolism, including HSL, CES1/2 and acyl-coenzyme A thioesterase (ACOT1) were detected. Within the 2-AG biosynthetic active enzymes, we again exclusively detected DAGLβ activity. Taken together, performing chemical proteomics allowed us to generate an overview of the lipolytic proteome of human placental tissue and demonstrated a broad spectrum of metabolic hydrolase activities.

As the human placenta is composed of different highly specialized cell types, we aimed to determine the localization of DAGLβ in situ, by using specific RNA probes. To localize the transcripts to distinct cell types of the placenta, immunofluorescence (IF) was applied. The visualization of DAGLβ transcripts was combined with cytokeratin 7 (CK7) staining, representing trophoblasts, which are building up the first structural barrier between maternal and fetal compartment and CD163 as a pan macrophage marker ( Figure 3A ). To identify feto-placental endothelial cells, which are lining placental vessels and are in direct contact with fetal blood, Von- Willebrand Factor (VWF) was applied ( Figure 3B ). We localized DAGLβ transcripts mainly to CK7 positive trophoblasts ( Figure 3A ). Quantitative analysis of the signals revealed that 54% of DAGLβ mRNA was localized to trophoblasts (T), while negligible signals of 2% and 3% were detected in endothelial cells (E) and macrophages (M), respectively ( Figure 3C ). In contrast to DAGLβ, we could not detect clear signals for DAGLα, confirming our observation that this enzyme is poorly expressed in the human placenta ( Supplementary Figure 2 ).

To better understand the specific function of DAGLβ in the intact organ, the lipid profile was examined after tissue perfusion with/out inhibitor. First, we examined the extent of DAGLβ inhibition obtained after perfusion by in-gel ABPP. Using DH379, we again identified the DAGLβ band (~70 kDa) in the vehicle perfused proteome ( Figure 4A ; Supplementary Figure 3 ). By co-application of DH376 [1µM], signal intensity was strongly reduced by 87%, confirming target engagement ( Figure 4B ). Thus, our results suggest that the application of DH376 at a concentration of 1 µM in an ex vivo setting leads to substantial inhibition of DAGL, which may also be accompanied by changes in the lipid profile. To examine the consequences of DAGLβ inhibition under the applied conditions, we analyzed changes in lipid species in perfused placental tissue. In accordance with the proposed cellular function of DAGL, total MAG levels were decreased by 60% ( Figure 4C ), resembled by a reduction of all detected MAG species ( Figure 4D ). Notably, the obtained data further demonstrated significant decreased 2-AG levels (MAG 20:4) upon inhibition of DAGLβ ( Figure 4D ). Interestingly, decreased MAG levels were not accompanied by augmented DAG concentrations, but a trend to decreased levels could be detected ( Figure 4E ). In fact, we observed significant reductions in specific DAG species, including DAG 32:0-16:0, DAG 34:2-18:2 and DAG 36:2-18:2 ( Figure 4F ). Furthermore, FA concentrations showed a downward trend in inhibitor perfused tissues ( Figure 4G ), of which eicosenoic acid (FA 20:1) was significantly reduced, indicating an effect on the hydrolysis of sn-1 FA of DAGs ( Figure 4H ). In contrast, FA levels in the maternal and fetal circuit remained unchanged ( Supplementary Figures 4A, B ).