The endocannabinoid 2-archidonoylglycerol (2-AG) is an important signaling lipid in the central nervous system (CNS). It acts as a retrograde messenger that activates the presynaptic cannabinoid receptor type 1 (CB1R), thereby regulating neurotransmitter release. 2-AG is involved in a variety of physiological processes, including modulation of memory, energy balance, and emotional states, such as stress and anxiety (Di Marzo, 2011). Biochemical, pharmacological and genetic studies have established diacylglycerol lipases α and β (DAGLα, DAGLβ) as the main biosynthetic enzymes that produce 2-AG in the brain by catalyzing the sn-1-specific hydrolysis of diacylglycerol (DAG) to generate 2-AG (Bisogno et al., 2003; Figure 1). For example, congenital deletion of DAGLα or DAGLβ resulted in 80 and 50% reduction, respectively, of brain 2-AG levels in knockout (KO) mice as compared to wild type (WT) littermates (Gao et al., 2010). Pharmacological studies with covalent, irreversible, dual DAGL inhibitors, such as DH376 and DO34, showed that acute blockade of 2-AG biosynthesis in the mouse brain reduced neuroinflammatory responses (Ogasawara et al., 2016), reversed LPS-induced anapyrexia (Ogasawara et al., 2016), reduced food intake (Deng et al., 2017), and modulated cocaine-seeking behavior (McReynolds et al., 2018) and stress responses (Bluett et al., 2017).

The life span of 2-AG signaling at the synapse is tightly controlled. Monoacylglycerol lipase (MAGL) (Dinh et al., 2002) and α,β-hydrolase domain containing protein 6 and 12 (ABHD6, ABHD12) have been identified as the key enzymes terminating the physiological role of 2-AG. They hydrolyze the ester bond in 2-AG, thereby generating glycerol and arachidonic acid (AA) (Blankman et al., 2007; Marrs et al., 2010; Figure 1). MAGL is the predominant lipase in the brain covering over 85% of 2-AG hydrolysis, whereas ABHD6 and ABHD12 account for 4 and 9%, respectively (Blankman et al., 2007).

2-Archidonoylglycerol signaling is not only important in the adult brain, multiple studies have also provided evidence of a functional role of 2-AG during neural developmental processes (Oudin et al., 2011b), including axonal growth and guidance (Bisogno et al., 2003; Harkany et al., 2007; Watson et al., 2008; Wu et al., 2010), differentiation (Jung et al., 2011), and neurogenesis (Gao et al., 2010; Oudin et al., 2011a). In adult mice, DAGLα is mainly restricted to postsynaptic sites on neurons, whereas DAGLβ is expressed in microglial cells. Importantly, DAGLα and DAGLβ are expressed by neurons at developing axonal tracts during neuronal development (Bisogno et al., 2003; Berghuis et al., 2007; Watson et al., 2008; Oudin et al., 2011a). Jung et al. (2011) have investigated the role of DAGLs in neuronal differentiation using retinoic acid (RA)-induced neurite outgrowth in murine neuroblastoma cell line Neuro-2a and found that RA elevated cellular 2-AG levels in Neuro-2a cells during differentiation and recombinant expression of DAGLα or DAGLβ increased neurite outgrowth, whereas silencing the expression of DAGLs using shRNAs reduced the number of cells with neurites. The contribution of the endogenously expressed DAGL enzymes to 2-AG biosynthesis in these cells is, however, less clear. Here, we sought to further test the role of the two DAGL isoforms in 2-AG biosynthesis in Neuro-2a using pharmacological, analytical, and genetic methods.

To investigate the role of DAGL in neuronal differentiation, Neuro-2a cells were incubated with RA, which induced a time-dependent outgrowth of neurites (Figures 2A,B). This resulted in increased 2-AG net levels as determined by liquid chromatography-mass spectrometry (LC-MS) (Figure 2C). These findings confirm and extend a previous study (Jung et al., 2011). Of note, anandamide (AEA) and AA levels were also significantly increased (Figure 2C). To check whether endogenously expressed DAGLs are responsible for the 2-AG production during differentiation, the cells were incubated with the dual DAGL inhibitor DH376. 2-AG and AA, but not AEA, levels were reduced by DH376 (Figure 2D). The inhibitor also impaired differentiation of Neuro-2a cells, as indicated by the reduced fraction of neurite bearing cells after 24, 48, and 72 h of RA stimulation (Figure 2B). This suggested that DAGL-dependent 2-AG and/or AA production plays a role in the differentiation process, as previously suggested (Jung et al., 2011).

To investigate which DAGL isoform is responsible for 2-AG production in Neuro-2a cells, we used a genetic approach as no subtype-specific DAGL inhibitors are currently available. Of note, single cell heterogeneity (in 2-AG production and neurite outgrowth) prevented the unequivocal analysis of single cell clone knockouts (Supplementary Figure S1). Therefore, disruption of DAGLα and DAGLβ genes was performed by three sequential rounds of transfection of Cas9 and single guide RNA’s (sgRNA) in Neuro-2a cell populations (Supplementary Figures S2A–C). This yielded three Neuro-2a knockdown (KD) populations: DAGLα KD, DAGLβ KD, and DAGLα-β KD. DAGLα and DAGLβ activity in these cell populations was measured using activity-based protein profiling (ABPP) to determine the efficiency of the genetic disruption. ABPP is a chemical proteomic method that uses chemical probes (e.g., fluorophosphonates (FP) or β-lactones) to assess the functional state of various enzymes, here DAGLα and DAGLβ, in native biological systems. When coupled to fluorescent reporter groups, activity-based probes (ABPs) enable visualization and quantification of enzymatic activity in complex proteomes by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and in-gel fluorescence scanning. When coupled to a biotin reporter group, ABPs enable affinity enrichment and identification of enzyme activities by mass spectrometry (MS)-based proteomics. Gel-based ABPP with a fluorescent FP-probe (FP-TAMRA) and β-lactone probe MB064 showed a reduction (>70%) of active DAGLβ in the DAGLβ KD and DAGLα-β KD populations (Figures 3A,B), without affecting other serine hydrolase activities. LC-MS-based chemical proteomics confirmed these findings (Figure 3D). Of note, no DAGLα activity was observed in either of the populations, including WT Neuro-2a (Figures 3A,B). The residual DAGLβ activity can be explained by a transfection efficiency below 100% and by insertion or deletion of a full codon upon Cas9-mediated DNA modification, thus preventing the frameshift that generally results in an early stop-codon.

Next, the 2-AG levels of the genetically modified Neuro-2a populations were quantified using LC-MS. Surprisingly, despite a >70% reduction in levels of active DAGLβ and no detectable DAGLα activity, the 2-AG levels in the DAGLα KD, DAGLβ KD or double DAGLα-β KD populations were not different from WT populations (Figure 3C). To test whether the 2-AG production in the KD populations was still sensitive to DH376 treatment, cells were incubated in situ (live cells) with DH376 (100 nM, 2 h). Lipidomics analysis on these samples revealed that DH376 again reduced cellular levels of both 2-AG and AA by 50% in all populations (Figure 3C). Of note, a small but significant increase in AEA levels was detected in the double DAGLα-β KD populations (Figure 3C), which could be due to increased levels of active ABHD4 (Figure 3D). Taken together, these data suggest that residual DAGLβ activity may be responsible for generating the entire pool of 2-AG, which seems unlikely, or that an alternative, unidentified enzyme, which is sensitive to DH376, contributes to 2-AG production in Neuro-2a cells.

To identify all 2-AG producing enzymes targeted by DH376 in Neuro-2a, we leveraged a chemical proteomics strategy in which the alkyne moiety of DH376 served as a ligation handle to introduce a reporter group via a copper(I)-catalyzed azide-alkyne cycloaddition (“click” chemistry) (Rostovtsev et al., 2002). Neuro-2a cells were incubated with DH376, lysed and the covalently bound inhibitor-target complexes were conjugated to Cy5-azide and visualized by SDS-PAGE and in-gel fluorescence scanning (Figures 4A,B). Apart from DAGLβ, one other fluorescent band with a molecular weight of ∼35 kDa was detected. Competitive ABPP using MB064 and FP-TAMRA suggested that this activity could be ABHD6, which was previously also reported as an off-target of DH376 (Ogasawara et al., 2016; van Rooden et al., 2018). To confirm the identity of this protein expressed by Neuro-2a cells, chemical proteomics was employed (Figure 4C). DAGLβ and ABHD6 were identified as the only targets of DH376 in Neuro-2a cells (Figure 4D). Thus, these data suggested that ABHD6 could be responsible for 2-AG production in conjunction with DAGLβ.

α,β-Hydrolase domain containing protein 6 is known to hydrolyze additional lipids than 2-AG, including lysophosphatidyl species (Thomas et al., 2013) and bis(monoacylglycero) phosphate (Pribasnig et al., 2015). To determine if ABHD6 can use DAG as a substrate, a DAG hydrolysis assay was developed based on fluorescent 1-nitrobenzoxadiazole-decanoyl-2-decanoyl-sn-glycerol (NBD-DAG) substrate. Lysates from HEK293-T cells overexpressing recombinant human ABHD6 (Figure 5A) or its catalytically inactive mutant (ABHD6^S148A) as a negative control, were incubated with NBD-DAG and analyzed by HPTLC. Lysates from HEK293-T cells expressing DAGLα or its catalytically inactive mutant (DAGLα^S472A) served as positive and negative controls, respectively (Supplementary Figure S3A). Both DAGLα and ABHD6 exhibited DAG-lipase activity as their overexpression resulted in the hydrolysis of NBD-DAG, whereas their mutants did not (Figures 5B,C and Supplementary Figures S3B,C), thereby showing that ABHD6 hydrolyzes the sn-1 ester bond of an sn-1-acyl-2-decanoyl-glycerol. Of note, the NBD-DAG hydrolysis measured in HEK293-T cells expressing GFP reflects the conversion of the substrate by endogenous hydrolases, including ABHD6, and to a lesser extent DAGLβ.

In view of the abundant ABHD6 activity in different brain regions (Baggelaar et al., 2017), we tested whether ABHD6 contributes to DAG hydrolysis in mouse brain. However, no reduction in DAG-NBD hydrolysis was observed in mouse brain lysates treated with the selective ABHD6 inhibitor KT182 or by comparing lysates from Abhd6^–/– mouse brain and WT controls (Supplementary Figure S4). Of note, mice could not be age-matched which may affect expression levels. However, aside from the loss of ABHD6 in knockout tissue, no evident discrepancies were observed in the lipase activity profiles (Supplementary Figure S4A).

Next, we determined if ABHD6 could also hydrolyze endogenous DAGs in intact cells. To this end, recombinant ABHD6 was overexpressed in HEK293-T cells (DAGLα as positive control and the catalytically inactive mutants as negative controls) and endogenous DAG (16:0, 20:4) levels were determined by targeted lipidomics (Figure 5D and Supplementary Figure S3D). Both ABHD6 and DAGLα overexpression reduced the levels of DAG (16:0, 20:4), whereas overexpression of the catalytically inactive mutants had no effect on the DAG-levels. DH376 treatment of the transfected cells prevented the reduction in DAG levels and in fact led to an increase in this lipid species, indicating that this DAG species serves also as an endogenous substrate for DAGLα/ABHD6. Overexpression of ABHD6 did not affect 2-AG levels, in line with the dual MAG/DAG-lipase character of ABHD6, whereas its final product AA was increased (Figure 5D). In summary, these data indicate that ABHD6 acts as a DAG/MAG-lipase using DAG (16:0, 20:4) as an endogenous substrate.

To check whether endogenous cellular levels of 2-AG and AA are controlled by ABHD6, a Neuro-2a ABHD6 KD population and a triple DAGLα-β-ABHD6 KD population were generated (Figure 6A and Supplementary Figures S2D–F). ABHD6 KD had no effects on 2-AG or AA levels (Figure 6C), which suggests that DAGLβ and ABHD6 activities can compensate for each other. The KD efficiency at the protein level was reduced in the triple DAGLα-DAGLβ-ABHD6 KD as compared to the single KDs as determined by gel-based ABPP (Figures 6A,B) and chemical proteomics (Figure 6D). We noticed that 40–50% of active ABHD6 and 30–40% active DAGLβ remained. In line with the reduced ABHD6 and DAGL activity in the triple KD, both 2-AG and AA levels were reduced by approximately 30%.

Finally, in light of the finding that ABHD6 can act as a DAG lipase, the DAGLβ and ABHD6 activity in Neuro-2a cells during RA-induced differentiation was mapped by gel-based ABPP (Figure 7A). A threefold increase in the amount of active ABHD6 was observed in differentiated Neuro-2a cells, whereas the amount of active DAGLβ was decreased (Figure 7B). However, inhibition of ABHD6 by the ABHD6 inhibitor KT182 did not abolish the 2-AG increase during RA-induced differentiation (Supplementary Figure S5). As the dual ABHD6/DAGL inhibitor DH376 inhibitor did reduce 2-AG levels and differentiation (Figure 2), these data suggest that the RA-induced 2-AG production (Figure 1B) in Neuro-2a cells is due to combined DAGLβ and ABHD6 activity.

In the present study, we extend our understanding of ABHD6 to a dual DAG/MAG-lipase that produces 2-AG. We also show that ABHD6 activity is involved in 2-AG production in Neuro-2a cells during RA-induced differentiation. Previously, it was shown that the RA-induced neurite outgrowth is a CB1R dependent process (Jung et al., 2011). The dual ABDH6/DAGL inhibitor DH376 reduced RA-induced Neuro-2a differentiation, thereby supporting the notion that 2-AG is involved in this process. Nevertheless, it is likely that other lipid mediators also play a role, because DH376 did not abolish neurite outgrowth and AEA and AA levels were dramatically increased during differentiation.

α,β-hydrolase domain containing protein 6 is a highly conserved metabolic hydrolase that belongs to the serine hydrolase family (Navia-Paldanius et al., 2012), mainly known as a 2-AG hydrolytic enzyme, but its substrates also include other monoacylglycerols (Navia-Paldanius et al., 2012), as well as various lysophosphatidyl species (Thomas et al., 2013) and bis(monoacylglycero)phosphate (Pribasnig et al., 2015). Using biochemical, genetic, and pharmacological approaches, we expanded the substrate scope of ABHD6 toward DAGs. Considering both MAG and DAG hydrolysis are mediated by the same catalytic serine (S148) (Figure 5; Navia-Paldanius et al., 2012), the respective MAG- and DAG-activities are therefore likely to be driven by relative substrate and product concentrations rather than by regulatory mechanisms. Our data indicate that 2-AG and AA are both produced by DAGL and ABHD6 during neuronal differentiation.

Although ABHD6 accounts for only 4% of 2-AG hydrolysis in mouse brain homogenates (Blankman et al., 2007), it is an important factor of MAG hydrolysis in specific cell- or tissues types, for example, in BV-2 cells where MAGL is lacking (Muccioli et al., 2007). In a similar fashion, the contribution of ABHD6 to DAG hydrolysis and 2-AG biosynthesis in the whole brain is likely to be limited, but it may be more pronounced in specific cell-types and physiological processes in the absence of DAGLα or DAGLβ. In the adult central nervous system, the postsynaptic localization of ABHD6 (Marrs et al., 2010) would indeed allow for retrograde endocannabinoid signaling. In the young brain, ABHD6 mRNA is found detected in progenitor cells (Gokce et al., 2016), which may provide an additional mechanism for cell-autonomous eCB-signaling (Goncalves et al., 2008). Notably, the localization of MAGL is complementary to that of ABHD6 (Straiker et al., 2009) with MAGL mainly localized at the presynaptic site. In a global MAGL KO mouse model no compensatory effects were observed by ABHD6 (Taschler et al., 2011). Similarly, ABHD6 deletion using antisense oligonucleotides did not affect MAGL activity in peripheral tissue (Thomas et al., 2013). Taken together, this suggests that, despite both being MAGLs, the physiological functions of ABHD6 and MAGL may be distinct.

Recent studies suggest that ABHD6 inhibitor have promising therapeutic efficacy in several preclinical mouse models of devastating diseases (Cao et al., 2019), such as metabolic syndrome (Thomas et al., 2013), chronic inflammation (Alhouayek et al., 2013), diabetes (Zhao et al., 2014), including traumatic brain injury (Tchantchou and Zhang, 2013), multiple sclerosis (Manterola et al., 2018), and epilepsy (Naydenov et al., 2014). Thus, it is important to further study the impact of the DAG lipase activity of ABHD6 on these pathophysiological processes to help develop novel therapeutics based on inhibiting ABHD6 activity.

We discovered that ABHD6 hydrolyzes DAG, thereby contributing to the production of the bioactive signaling lipid 2-AG during RA-induced differentiation of Neuro-2a cells. Our finding is supported by biochemical, genetic and pharmacological evidence. First, genetic knockdown of both DAGLα and DAGLβ in Neuro-2a cells had no effect on cellular 2-AG levels. Second, the dual DAGL/ABHD6 inhibitor DH376 abolished cellular 2-AG levels. Third, ABHD6 catalyzed the degradation of a fluorescent DAG substrate. Fourth, overexpression of ABHD6, but not its catalytically inactive mutant, reduced endogenous DAG (16:0, 20:4) levels, which was inhibited by DH376. Fifth, cellular 2-AG levels decreased upon triple KD of DAGLα-β-ABHD6 in Neuro-2a. These data extend the role of ABHD6 from a MAG lipase to a DAG lipase and suggest a potential role for ABHD6 in neuronal differentiation.

The raw data supporting the conclusion of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

The animal study was reviewed and approved by the mice were housed in a pathogenic-free facility in accordance with the National Institutes of Health; the Institutional Animal Care and Use Committee at the University of Washington approved all experiments.

AE and MS contributed to the conception and design of the study. AE carried out the experiments and analyzed the data. XD and VK measured and analyzed the lipidomics experiments. AB and BF assisted with the LC/MS measurements and data analysis of chemical proteomic experiments. TW assisted in sgRNA design and cloning. TW and AS assisted in experiments. NS and SS provided the mouse brain tissues from WT and Abhd6^–/– mice. HD cloned and prepared the plasmids for recombinant expression. AE and MS wrote the manuscript. All authors contributed to the manuscript revision, read, and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.