There are an estimated 500 million obese people worldwide (1). This obesity epidemic is a major contributor to the dramatic increase in prevalence of type-2 diabetes (T2D) and the rise in diabetes complications (2). Unfortunately, efforts to induce weight loss by encouraging changes in dietary intake and physical activity alone have not reversed the obesity epidemic; this suggests that combining these behavioral modifications with a safe and effective anti-obesity drug regimen is the approach most likely to result in meaningful and sustained loss of body fat (3). Although three new anti-obesity drugs were recently approved in the United States, each has significant safety issues that are being evaluated in post-marketing studies (3). There is a need for additional anti-obesity drugs to complement and/or replace those that are currently approved.

The endocannabinoid (EC) system has long been linked to the regulation of food intake and energy balance (4). The primary psychoactive cannabinoid, Δ⁹-tetrahydrocannabinol (THC), signals through the G-protein-coupled receptors Cnr1 and Cnr2, also known as CB1 and CB2, respectively (5). The two major endogenous ligands or ECs that signal through these receptors are 2-arachidonoylglycerol (2-AG) and N-arachidonoylethanolamine (AEA or anandamide). The final enzymatic step in the synthesis of 2-AG is catalyzed by either diacylglycerol lipase α or β (Dagla or Daglb); by contrast, a number of enzymes, including N-acylphosphatidylethanolamine-hydrolyzing phospholipase D (Napepld), catalyze the final step in the synthesis of AEA (6–9). Cnr1 is highly and widely expressed in central nervous system (CNS) neurons, including neurons that regulate energy balance, and is also expressed in peripheral organs important for metabolic control; by contrast, Cnr2 is expressed primarily by cells involved in regulating immune function (10). Dagla is much more highly expressed in CNS neurons than in peripheral tissues with CNS expression mirroring that of Cnr1, Daglb is widely expressed but with much lower CNS expression than Dagla, and Napepld is widely expressed with highest expression in the CNS (6, 11). In the CNS, Dagla and Daglb demonstrate dendritic expression, consistent with the model that ECs, and particularly 2-AG, produced by post-synaptic cells inhibit presynaptic release of excitatory or inhibitory neurotransmitters by signaling through Cnr1 on presynaptic terminals (6, 12).

Food intake can be stimulated by oral delivery of THC, and by injections of either 2-AG or AEA, and this effect is blocked by the Cnr1 inverse agonist rimonabant (13–16). Further evidence that Cnr1 mediates this feeding response is provided by studies showing that hypophagia and decreased body weight (BW) are found in Cnr1 knockout (KO) mice and in mice chronically treated with rimonabant (13, 17–21). Long-term rimonabant treatment induced weight loss not only in rodent models but also in obese humans studied in multiple clinical trials (22–24). Unfortunately, neuropsychiatric side-effects, which were likely on-target considering the role of ECs in neural pathways regulating a wide range of emotional behaviors, ultimately led to withdrawal of rimonabant from the market (24–26).

In an effort to identify potential drug targets, Lexicon Pharmaceuticals performed a high-throughput phenotypic screen on >4,650 KOs of independent genes encoding druggable proteins (27–31). As part of this phenotypic screen, body fat was measured in cohorts of chow-fed mice from 3,651 KO lines. Among KO lines identified with potential body fat phenotypes, Cnr1 KO mice had a lean phenotype, confirmed using additional cohorts of mice (30), that was consistent with the lean phenotype of Cnr1 KO mice reported by others (17, 20). Although Dagla, Daglb, and Napepld KO lines have been generated (9, 32, 33), their body fat levels have not been reported. Because these three proteins are potentially druggable enzymes, Dagla, Daglb, and Napepld KO lines were phenotyped in our screen. We chose to review the screen results for body fat from these KO lines to determine if any or all shared the lean phenotype of Cnr1 KO mice, and if so to examine them in greater detail to determine how closely other phenotypic characteristics mirrored those of Cnr1 KO mice, including behavioral phenotypes that suggest the possibility of on-target neuropsychiatric side-effects which would likely preclude developing inhibitors of these enzymes as anti-obesity drugs for humans.

Similar to Cnr1 KO mice, Dagla KO mice were lean relative to WT littermates in our high-throughput phenotypic screen of 3,651 KO lines maintained on chow diet; by contrast, neither Daglb nor Napepld KO mice were lean in this screen (Figure 1A). Dagla mice had normal Mendelian distribution at weaning (923 WT, 1882 Het, 971 KO) and appeared healthy, similar to Cnr1 (378 WT, 717 Het, 318 KO), Daglb (173 WT, 332 Het, 169 KO), and Napepld (77 WT, 150 Het, 82 KO) mice. To confirm the lean phenotype observed during screening, additional cohorts of chow-fed male and female Dagla KO mice were studied and were found to have significantly decreased BW, body fat, and lean body mass (LBM) at weaning and through 20 weeks of age (Figures 1B,C).

Dagla KO mice fed HFD from weaning developed significantly decreased BW, body fat, and LBM relative to WT littermates (Figures 2A,B). When we combined QMR data obtained at 15–17 weeks of age from multiple chow- and HFD-fed cohorts, we found that BW, body fat, and LBM were significantly decreased by 21, 47, and 11%, respectively, in 131 Dagla KO mice compared to 169 WT littermates (Figures 2C–E; Table S1 in Supplementary Material); because normalized data from male and female KO mice were comparable, they were combined for this analysis. The lean phenotype persisted with age, and was still present in Dagla KO mice aged more than 1 year (Table 1). The body composition phenotype of Dagla KO mice was confirmed by DEXA analysis (Table S2 in Supplementary Material), which also revealed a slightly decreased bone mass for 29-week-old Dagla KO mice that was probably appropriate for their slightly smaller skeletal frame as demonstrated by femur length measurements (Figure 2F). The lean phenotype of Dagla KO mice is quite similar to that observed in our Cnr1 KO mice (30). In an update of our Cnr1 KO data, we combined QMR measurements obtained at 16–32 weeks of age from multiple chow- and HFD-fed cohorts and found that BW, body fat, and LBM were significantly decreased by 18, 45, and 8%, respectively, in 80 Cnr1 KO mice compared to 102 WT littermates (Figures 3A–C; Table S3 in Supplementary Material); because normalized data from these male and female KO mice were comparable, they were combined for this analysis. By contrast, neither Napepld nor Daglb KO mice raised on HFD showed decreased body fat at 15 weeks of age (Figures 3D,E). In addition, DAGLa/b DKO mice showed no enhancement of the low body fat phenotype over that observed in DAGLa KO mice (Figure 3F); these mice appeared healthy at weaning, and their Mendelian distribution was normal.

We sought to identify the mechanism behind the low-body fat phenotype of Dagla KO mice. We found that Dagla KO mice ate significantly less than WT littermates at weaning (Figure 4A), similar to Cnr1 KO mice (Figure 4B). A pair-feeding study, performed on female Dagla mice beginning at weaning, showed that pair-fed WT mice had BWs similar to Dagla KO mice and significantly less than WT mice fed ad lib (Figure 4C), suggesting that the lean phenotype of Dagla KO mice at weaning was primarily due to decreased food intake. By contrast, when we studied 27-week-old Dagla KO mice with an established lean phenotype, we found that WT littermates pair-fed to these Dagla KO mice showed an insignificant loss of BW relative to WT littermates that were fed ad lib (Figure 4D). Adult Dagla KO mice did not show evidence of malabsorption (Figures 5A,B) and, in addition, had VO₂ (Figure 5C) and total activity (Figure 5D) levels that were not different from those of WT littermate controls. The total activity data were confirmed using a different technology in a separate cohort (Figure 5E). Similar to Dagla KO mice, Cnr1 KO mice had total activity levels that were comparable to levels of WT littermates (Figure 5F).

We evaluated the effect of Dagla deficiency on metabolic parameters. Although Dagla KO mice maintained on HFD had fasting blood glucose levels that were not different from WT, and responded to an OGTT with only modest lowering of glucose excursions relative to WT values, they did exhibit significantly lower 0 and 30 min insulin levels (Figures 6A–D); thus, less insulin was required to maintain post-prandial glucose excursions in these non-diabetic mice. Cnr1 KO mice maintained on HFD also showed no difference from WT mice in terms of fasting blood glucose and, during OGTTs, exhibited an insignificant improvement in glucose excursions associated with significantly lower 0 and 30 min insulin levels (Figures 6E,F). Fasting TG and total cholesterol levels tended to be lower in multiple cohorts of Dagla KO mice-fed HFD (Table S4 in Supplementary Material); when data were normalized such that mean WT values for each male and female cohort were assigned a value of 100%, and all data were then combined, TG and total cholesterol levels were significantly lower in Dagla KO mice, measuring 75 and 87% of WT littermate values, respectively (Figure 6G). Cnr1 KO mice weaned onto HFD showed a similar pattern (Table S4 in Supplementary Material); when data were again normalized and combined, TG and total cholesterol levels were significantly lower in Cnr1 KO mice, measuring 83 and 80% of WT littermate values, respectively (Figure 6H). We also studied whether Dagla deficiency would protect against hepatic steatosis and inflammation by placing Dagla KO mice, ApoE KO mice, Dagla/ApoE DKO mice, and WT littermates on Western diet from weaning (Table 2). When these mice were studied at 20–24 weeks of age, Dagla but not ApoE KO mice were lean relative to WT littermates, and histologic examination of hepatic tissue demonstrated that Dagla deficiency significantly protected WT mice from steatosis but not inflammation, whereas ApoE deficiency had no effect on either parameter in this analysis. Serum analysis showed that ApoE KO mice had increased TG, total cholesterol, and AST relative to WT littermates, and that the addition of Dagla deficiency to ApoE deficiency significantly lowered each of these parameters. Comparable studies were not performed using Cnr1 KO mice.

Dagla and Cnr1 KO mice were also evaluated for their performance during a number of behavioral tests (Table 3). During the tail suspension test, neither Dagla nor Cnr1 KO mice showed a significant difference in immobility time relative to WT littermates. The forced swim test showed significantly decreased immobility time in Dagla KO mice, while a much smaller cohort of Cnr1 KO mice showed a strong trend (P = 0.07) toward a similar decrease. In the open-field test, neither Dagla nor Cnr1 KO mice differed from WT littermates in total distance traveled, and both showed significantly less rearing activity than did their WT controls; by contrast, the Cnr1 KO mice spent significantly less time than WT littermates in the center of the open field, whereas Dagla KO and WT littermates were not different for this parameter. This last finding was revisited in the platform test, which provides a more specific measure of anxiety-related behavior (38); Dagla KO mice spent significantly more time than WT mice in the lighted area, and a much smaller cohort of Cnr1 KO and WT mice did not differ significantly for this parameter. Finally, both Dagla and Cnr1 KO mice buried significantly fewer marbles during the marble burying test, and both took significantly longer than their WT littermate controls to respond to a thermal stimulus during the hot plate test. For all tests except the marble burying in Dagla KO mice, and the forced swim and platform tests in Cnr1 KO mice, the findings in Table 3 were confirmed with at least one additional independent cohort of mice (data not shown).

Long-term survival was significantly decreased in Dagla KO mice (Figure 7). Despite this impaired survival, aging Dagla KO mice appeared healthy. In a cohort of 12 Dagla KO mice that were at least 1 year of age at necropsy, no pathological changes were noted in brain, heart, or other tissues examined that could explain the decreased survival of these mice (data not shown). On three separate occasions, Dagla KO mice undergoing QMR were observed to have spontaneous seizure activity and none of these mice survived the seizure episode, suggesting that the decreased survival was related to an underlying neurological abnormality which was not evident during histopathologic analysis of brain tissue.

Our Dagla and Cnr1 KO mice share the same lean phenotype when raised on either chow or HFD. In fact, the similarity of their lean phenotypes was remarkable, with body fat 47 and 45% lower in Dagla and Cnr1 KO mice, respectively, relative to WT littermates. By contrast, Daglb and Napepld KO mice had normal body fat stores. The lean phenotype of our Cnr1 KO mice confirms published data (17, 20, 30). By contrast, body fat data have not been reported for Dagla or Daglb KO mice, although one study did report that BW was significantly decreased in Dagla KO mice and slightly decreased in Daglb KO mice (32). The absence of a body composition phenotype in Napepld KO mice is consistent with recent data indicating that this enzyme is not required for maintenance of normal CNS AEA levels (9). We also found that Dagla and Cnr1 KO mice are hypophagic, consistent with past studies showing decreased feeding behavior in Cnr1 KO mice (17, 18, 20). In our studies, decreased food intake was demonstrated in weanling Dagla and Cnr1 KO mice at a time when their lean phenotype was developing. In addition, a pair-feeding study performed in weanling Dagla KO mice showed that hypophagia was the mechanism behind their lean phenotype, a result identical to that reported for a pair-feeding study of weanling Cnr1 KO mice (17). The important role of hypophagia in development of the lean phenotype in Dagla and Cnr1 KO mice suggests that the lean phenotype results primarily from loss of Dagla-generated 2-AG signaling through Cnr1 expressed by CNS rather than by peripheral tissues. This hypothesis is consistent with the observations that: (1) Cnr1 and Dagla are most highly expressed in the CNS, (2) brain 2-AG levels are much lower in Dagla KO mice compared to Daglb KO mice, and (3) removal of Cnr1 from neurons known to regulate energy balance, but not from non-neuronal peripheral tissues, is all that is required to reproduce the lean phenotype observed in global Cnr1 KO mice [(32, 33, 42, 43); our unpublished observations]. Interestingly, hypophagia was not as pronounced in older Dagla KO mice, and in fact pair-feeding of WT mice to Dagla KO mice was not associated with marked loss of BW relative to ad lib-fed WT littermates, a finding that exactly reproduced results reported for pair-feeding studies using older Cnr1 KO mice (17). Such results may simply reflect increased food intake as an adaptation to maintain already depleted body fat stores and LBM in adult Dagla and Cnr1 KO mice, or they could suggest that food intake-independent mechanisms were operative in adult KO mice. However, activity was not increased in either our adult Dagla or Cnr1 KO mice in three independent assays and there was no evidence of malabsorption in our Dagla KO mice. Also, VO₂ was not increased in our Dagla KO mice or in Cnr1 KO mice studied by others (17); however, the error inherent in the indirect calorimetry method and the ability of Cnr1 inverse agonists to increase energy expenditure in rodents and humans (44, 45) suggest that additional studies are needed to determine if increased energy expenditure contributes to the lean phenotype of older Dagla KO mice.

A number of metabolic parameters were studied in Dagla and Cnr1 KO mice. Neither HFD-fed Dagla nor Cnr1 KO mice showed consistent improvements in fasting blood glucose levels or glucose excursions during OGTTs, but both KOs had significantly lower insulin levels during a fast and 30 min after glucose challenge, along with lower triglyceride (TG) and total cholesterol levels. These results are similar to Cnr1 KO mouse data published by others (20, 42, 43, 46). Dagla KO mice-fed western diet also showed a decrease in hepatic steatosis, which mirrored their decrease in total body fat. Past studies initially suggested that hepatic Cnr1 deficiency alone prevented the HFD-induced development of hepatic steatosis, insulin resistance, and dyslipidemia, whereas extra-hepatic Cnr1 deficiency was required to prevent HFD-induced obesity (46). Although subsequent studies reported that CNS-specific Cnr1 deficiency alone prevented HFD-induced hepatic steatosis, insulin resistance, and dyslipidemia (42, 43), recent data indicate that peripherally restricted Cnr1 antagonists can indeed prevent this metabolic syndrome (47). Our data suggest that, wherever the Cnr1 that confers each of these effects is expressed, Dagla must also be expressed to provide the 2-AG that serves as the EC signal.

Mice treated with rimonabant show the same resistance to hepatic steatosis, insulin resistance, dyslipidemia, and obesity as do mice lacking Cnr1 (19, 20, 42, 43, 46, 48) or Dagla (data presented here), suggesting that Dagla inhibitors may achieve the same effects as Cnr1 inverse agonists if delivered to the right location. In obese humans with T2D, rimonabant significantly lowered BW, A1C, insulin resistance, and serum TGs (22, 49), suggesting that Dagla inhibitors will also be effective in these individuals. Of the few Dagla inhibitors studied, the most promising data are for O-7460, a small molecule that lowered food intake in a dose-dependent manner during the 14 h after a single intraperitoneal injection; the highest dose was associated with slight but significant decreases in BW and hypothalamic 2-AG levels (50). Unfortunately, hypophagia is a common sign of off-target toxicity for many compounds. For this reason, developers of the Cnr1 inverse agonists, rimonabant and taranabant, demonstrated that their compounds lowered food intake in WT but not Cnr1 KO mice (19, 51). Dagla inhibitors should also be tested in Dagla KO mice early in preclinical development to confirm that their effects on food intake are truly on-target.

The EC system plays an important role in regulating emotional behavior (26). Although the Cnr1 inverse agonist rimonabant was an effective anti-obesity agent, it was often associated with affective disturbances such as anxiety and depression that made it unsuitable for routine use (22, 24, 25). The value of Dagla as an anti-obesity target would be enhanced if the favorable metabolic phenotype shared by Dagla and Cnr1 KO mice was not linked to the undesirable neuropsychiatric effects associated with rimonabant. For this reason, we chose to evaluate Dagla and Cnr1 KO mice in a series of tests designed to evaluate anxiety, depression, and other behaviors. Interestingly, published studies that used these behavioral tests to evaluate Cnr1 KO mice, or WT mice treated with compounds that modulate Cnr1 signaling, reported results that were often inconsistent with each other and with data we collected on our Cnr1 KO mice (39, 40, 52–61). This suggested that the value in our behavioral screen would come from directly comparing the behavioral phenotypes of Dagla and Cnr1 KO mice obtained in our laboratory. We found that Dagla and Cnr1 KO mice differed only in the open-field time in center and the platform tests, two related tests that use the same open-field chamber; here Dagla KO mice exhibited a more anxiolytic behavior. Dagla and Cnr1 KO mice showed similar behaviors in the hot plate, marble burying, open-field rearing, forced swim, open-field distance traveled, and tail suspension tests; for the first four of these tests, the behavior of KO mice from both lines appeared different from WT. The fact that both Dagla and Cnr1 KO mice responded similarly in most of the tests suggests that the 2-AG synthesized by Dagla is mediating not only effects on pathways regulating food intake and various metabolic parameters but also effects on pathways that influence important neuropsychiatric behaviors. Our studies were limited by the fact that we only employed a few acute assays to assess depression and anxiety (62). In addition, the open-field time in center and platform tests suggest the possibility that the two KO lines may differ in anxiety-related behavior, with Dagla KO mice exhibiting a more anxiolytic response; this is a potentially important finding, because it raises the possibility that Dagla KO mice do not share an undesirable behavioral phenotype present in Cnr1 KO mice. Additional studies are therefore needed; if they ultimately confirm our overall impression that Dagla and Cnr1 KO mice share behavioral phenotypes, the response to Dagla inhibitors may well be similar to the response to Cnr1 antagonists and inverse agonists for both metabolic and neuropsychiatric effects. Finally, we observed seizure activity followed by death in three Dagla KO mice; this may explain the decreased survival of our Dagla KO line, which occurs despite a lack of histopathologic findings. Although we did not assess survival of our Cnr1 KO mice, others reported increased seizure activity and decreased survival without an obvious cause of death in Cnr1 KO mice (21). These shared findings provide further evidence for a close interaction between Dagla and Cnr1 in the normal regulation of multiple neurologic pathways.

Our high-throughput phenotypic screen correctly identified the lean phenotypes of Cnr1 and Dagla KO mice, just as it correctly identified the body fat phenotype of many KO lines in past studies; some of these KOs share their phenotype with humans having inactivating mutations in the same gene (30, 35, 63). Our screen has also identified genes that play a key role in regulating bone mass (31). Continued mining of our KO data, and KO data generated by the International Mouse Knockout Consortium, should identify genes that, when mutated, (1) result in specific diseases and (2) can be targeted to develop therapeutics for obesity, diabetes, osteoporosis, and other indications.

In summary, our data suggest that, in pathways where ECs signal through Cnr1 to regulate food intake, certain other behaviors and also other aspects of the metabolic phenotype observed in Cnr1 KO mice, Dagla alone provides the 2-AG that serves as the EC signal. Based on these data, small molecule Dagla inhibitors with a pharmacokinetic profile similar to that of Cnr1 inverse agonists are likely to mirror the ability of these inverse agonists to decrease food intake, BW, TGs, and insulin resistance while improving glycemic control in obese patients with diabetes. Whether Dagla inhibitors will also be associated with undesirable neuropsychiatric events similar to those observed with use of Cnr1 inverse agonists, which must be considered possible based on data presented here, can only be answered with human clinical trials.

DP, KS, TL, ZD, UD, and BZ made substantial contributions to the conception or design of the work. DP, KS, UD, JG, NW, DD, RR, PV, GH, RB, and ZD made substantial contributions to the acquisition, analysis, or interpretation of data for the work. DP, ZD, and UD participated in drafting the work. JG, NW, DD, KS, TL, RR, PV, GH, RB, and BZ participated in revising it critically for important intellectual content. All authors provided final approval of the version to be published. All authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

All authors were employed by Lexicon Pharmaceuticals, Inc. at the time these studies were performed, and received compensation in the form of salary and stock options.

The Supplementary Material for this article can be found online at http://www.frontiersin.org/Journal/10.3389/fendo.2015.00086/abstract