Monoacylglycerol lipase – a target for drug development?

Introduction

Since antiquity, cannabis extracts have been used both recreationally and for therapeutic purposes. The influential physician Sir John Russell Reynolds, for example, wrote in 1890 that ‘In almost all painful maladies I have found Indian hemp by far the most useful of drugs’¹ (Reynolds, 1890; for a comprehensive review over the history of Cannabis sativa, see Mechoulam, 1986). The main psychoactive component of cannabis, Δ^{-tetrahydrocannabinol (Δ-THC) produces most of its effects in the body as a result of the activation of two G-protein-coupled receptors, termed CB}₁ and CB₂ receptors. Originally, CB₁ receptors were described as being primarily neuronal in localization, while CB₂ receptors were found mainly in cells of immune origin. However, more recent data has suggested that this is an oversimplification, and that both non-neuronal CB₁ receptors are found as well as a limited central nerve system distribution of CB₂ receptors (Atwood and Mackie, 2010). There is evidence for additional CB-like receptors, but currently, the evidence is not considered sufficiently strong for them to be formally included in the CB receptor family (Pertwee et al., 2010). Some of the physiological processes affected or regulated by CB receptors are shown in Figure 1.

Figure 1

(Patho)physiological processes affected or regulated by CB receptors in the body. For authoritative reviews on these areas, the reader is referred to Bab and Zimmer, 2008; Flygare and Sander, 2008; Katona and Freund, 2008; Yazulla, 2008; Guindon and Hohmann, 2009; Hiley, 2009; Maccarrone, 2009; Idris and Ralston, 2010; Basu and Dittel, 2011; Keimpema et al., 2011; Quarta et al., 2011. It is to be hoped that the ballet dancer shown in silhouette in the figure will recover from the pathologies in question.

The demonstration of CB receptors in the body led to the search and thereafter identification of a number of endogenous ligands, mainly (but not exclusively) arachidonic acid derivatives (Devane et al., 1992; Hanus et al., 1993; 2001; Mechoulam et al., 1995; Sugiura et al., 1995). As with the receptors, it is prudent to be restrictive in defining compounds as endogenous cannabinoids or endocannabinoids (eCBs), and in this review, the eCB system is defined as comprising the endogenous ligands 2-arachidonoylglycerol (2-AG, structure, see Figure 2) and anandamide (arachidonoylethanolamide, AEA; structure, see Figure 2), their target CB₁ and CB₂ receptors, and their primary synthetic and degradative enzymes. The life cycle of 2-AG is described in Box 1.

Figure 2

Chemical structures of the main eCBs and of selected MGL inhibitors. The two selective FAAH inhibitors discussed in the review are also shown to aid the reader.

Monoacylglycerol lipase (MGL) is a 33 kDa serine hydrolase that catalyses the hydrolysis of monoacylglycerols to their corresponding fatty acids (Karlsson et al., 1997). The enzyme is found both in the brain and in peripheral tissues such as the kidney, ovary, testis, adrenal gland, adipose tissue and heart (Karlsson et al., 1997). The three-dimensional structure of MGL has been elucidated by X-ray crystallography, and the enzyme has been shown to be a dimeric molecule with amphitropic properties, that is, it can exist both in a soluble form and associated with the membrane lipid bilayers (Bertrand et al., 2010; Labar et al., 2010). Mutagenesis studies have demonstrated the importance of the catalytic triad of ser^{, asp and his that the enzyme shares with other enzymes in the α/β hydrolase superfamily (}Karlsson et al., 1997). The substrate is recruited via a wide hydrophobic tunnel, where it can then interact with the catalytic triad at the end of the tunnel (Bertrand et al., 2010; Labar et al., 2010).

Although MGL can hydrolyse monoacylglycerols with a wide range of acyl chain lengths and degrees of unsaturation (Tornqvist et al., 1974; Nomura et al., 2010), most current interest has emanated from the finding that in the brain, MGL is highly expressed in regions with high CB₁ receptor densities (Dinh et al., 2002) and that in the mouse brain, it is responsible for the bulk (∼85%) of the metabolism of 2-AG (Blankman et al., 2007; for details concerning the other metabolic pathways for 2-AG, see Box 1).

Given the roles of the eCB system in the regulation of many processes including pain, appetite and even the control of cancer (Figure 1), enzymes targeting eCB catabolism are potentially important targets for pharmaceutical development. Currently, the main focus has been upon fatty acid amide hydrolase (FAAH), which primarily hydrolyses AEA and related N-acylethanolamines to their corresponding fatty acids. FAAH inhibitors are in pharmaceutical development, with pain as a major indication (Ahn et al., 2009). MGL inhibitors have lagged behind, but have a potential advantage over FAAH inhibitors in terms of eCB specificity of action given that the N-acylethanolamines are a class of compounds with multiple biological actions (De Petrocellis and Di Marzo, 2009; Hansen, 2010). In this review, the current state of the art with respect to (i) selective MGL inhibitors; (ii) the physiological role of MGL and (iii) the therapeutic potential of MGL inhibitors will be discussed.

Selective MGL inhibitors

MGL is inhibited by a variety of sulfhydryl reagents and disulphide compounds, mercuric compounds, fatty acid trifluoromethylketones and organophosphorus compounds such as methyl arachidonyl fluorophosphonate (MAFP; examples of structures, see Figure 2; Tornqvist and Belfrage, 1976; Di Marzo et al., 1999; Goparaju et al., 1999; Dinh et al., 2002; Ghafouri et al., 2004; Saario et al., 2005; Nithipatikom et al., 2005; Quistad et al., 2006; Labar et al., 2007; Tarzia et al., 2007; Nomura et al., 2008; Matuszak et al., 2009) although the selectivity and/or ‘drugability’ of these compounds is very limited. Nonetheless, an organophosphorus compound was used in vivo to demonstrate that reduction of the 2-AG hydrolytic capacity of the brain is associated with an increased level of 2-AG, with a corresponding decrease in arachidonic acid levels (Nomura et al., 2008). The sulfhydryl reagent N-arachidonoylmaleimide (structure, see Figure 2) has also been shown in vivo to potentiate the CB₁ receptor-mediated behavioural effects of 2-AG in the tetrad test of cannabinoid function (Burston et al., 2008). Early attempts to design substrate analogues with marked selectivity towards MGL over FAAH (the enzyme responsible for the hydrolysis of the other main eCB AEA) were unsuccessful (Ghafouri et al., 2004; Vandevoorde et al., 2005; Cisneros et al., 2007). A more recent substrate analogue [with respect to 2-oleoylglycerol (2-OG)] (2S,9Z)-octadec-9-ene-1,2-diamine, produced 50% versus 5% inhibition of MGL and FAAH respectively at a concentration of 100 µM (Magrioti et al., 2008). The compound showed analgesic activity in a model of visceral pain, but the potency of the compound towards MGL (K_i value 21.8 µM, Magrioti et al., 2008) is a limiting factor.

The first compound to be reported with purported selectivity for MGL over FAAH was a carbamate compound, URB602 (structure, Figure 2; Hohmann et al., 2005), although this in vitro selectivity has been contested (Vandevoorde et al., 2007). The compound, however, has functional selectivity since its administration increases levels of 2-AG but not of AEA in rat forebrain organotypic slice cultures (Hohmann et al., 2005). One possible explanation of this behaviour is that the degree of inhibition of MGL required to increase 2-AG levels is less than that needed towards FAAH to increase AEA levels. There is evidence supporting this assertion in the case of JZL184 (Long et al., 2009b, compound described in detail below; structure, see Figure 2): at a dose of 4 mg·kg ^{i.p. of the compound, a 75% inhibition of mouse brain MGL is seen and brain 2-AG levels are increased dramatically 4 h after dosing (although the levels of the oleoyl- and palmitoyl-homologues did not change). In contrast, at a dose of 40 mg·kg of the compound, which produced an even larger blockade (∼90% inhibition) of FAAH, no increase in AEA levels are seen, although levels of the AEA homologues oleoylethanolamide and palmitoylethanolamide increase by a factor of two (}Long et al., 2009b). Regardless as to how URB602 produces its selective increase in 2-AG brain levels, the low potency of the compound is a limitation, although it has been used in vivo (Table 1). Nonetheless, the compound has served as an inspiration for the identification and/or design of MGL (and MGL/FAAH) inhibitors (King et al., 2007; Muccioli et al., 2008; Long et al., 2009b; 2010; Bertrand et al., 2010; Szabo et al., 2011; Cisneros et al., 2012). The observation that sulfhydryl reagents are potent MGL inhibitors has also led to identification of octhilinone and N-octylbenzisothiazolinone, which inhibit MGL with approximately 20-fold selectivity over FAAH (King et al., 2009a; Matuszak et al., 2011). Replacement of the acyl side chain of the latter compound with the biphenyl moiety found in URB602 resulted in a compound (2-(4-methylbiphenyl)benzo[d]isothiazol-3(2H)-one) with 73-fold selectivity for MGL over FAAH (Matuszak et al., 2011).

Over the last few years, MGL inhibitors with novel structures have been identified. Starting from the diacylglycerol lipase α (DGLα)/broad specificity lipase inhibitor tetrahydrolipstatin, detailed structure activity relationship studies led to the identification of OMDM169 (structure see Figure 2) as an apparently competitive inhibitor of MGL with IC₅₀ values for the inhibition of 2-AG hydrolysis varying from 0.13 µM (rat cerebellar cytosolic fraction) to 10 µM (mouse paw skin). In comparisons for enzymes from the same source, OMDM169 had approximately 10-fold selectivity versus FAAH (rat brain) and versus DGLα (recombinant enzymes expressed in COS-7 cells). However, the compound inhibited recombinant pancreatic lipase with an IC₅₀ value of 0.6 µM (Ortar et al., 2008; Bisogno et al., 2009). The thiazolidinedione compound troglitazone, a peroxisome proliferator-activated γ ligand, was found to inhibit MGL with good selectivity versus FAAH, but the potency towards MGL was found to be highly dependent upon the assay used to assess the inhibition (Björklund et al., 2010). Following a screen of a commercial library of compounds, the naturally occurring terpenoid pristimerin (structure, see Figure 2) was identified as a potent reversible inhibitor of MGL (IC₅₀ value 93–400 nM, depending upon the enzyme source) with no effect upon FAAH. The compound also inhibited α/β-hydrolase domain 6 (ABHD6, IC₅₀ 98 ± 8 nM), and the combined effect upon this enzyme and MGL resulted in increased levels of 2-AG, but not the FAAH substrate palmitoylethanolamide, in rat cortical neurons in primary culture (King et al., 2009b). Whether or not the potentiation of 2-AG signalling contributes to the ability of pristimerin to produce apoptosis in breast and prostate cancer cells (Yang et al., 2008a; see below for discussion of the control of cancer by the eCB system) awaits elucidation. Be that as it may, the reversibility of this compound combined with its good selectivity versus FAAH is a potentially important property, and it is to be hoped that in vivo studies with this compound will be forthcoming.

The most extensively studied selective MGL inhibitor is JZL184 (structure, see Figure 2). This compound was designed by an activity-based protein profiling screen of a library of carbamate compounds, followed by chemical optimization of the best compound. JZL184 acts as a potent irreversible inhibitor of MGL with an approximate 300-fold selectivity versus FAAH (see Box 2 for details), and with an estimated brain t½ in the mouse of 7 h (Long et al., 2009b). In vivo, a dose of 16 mg·kg ^{i.p. given to mice produces almost complete inhibition of MGL with only partial inhibition of FAAH and no obvious effects on any of the other serine hydrolases investigated using the activity-based profiling technique (}Long et al., 2009b). The compound dose dependently increases 2-AG levels in the brain while AEA levels are not affected. At a dose of 16 mg·kg^{, this increase is seen at the earliest time point measured (0.5 h after dosage) and continues up to 8 h, having returned to baseline at 24 h after dosing (}Long et al., 2009b). However, repeated administration to mice of a high dose of JZL184 (40 mg·kg^{) results in a down-regulation of CB}₁ receptors and a tolerance to both the effects of the compound itself and of exogenous cannabinoids in vivo, a situation also seen when the enzyme is genetically deleted (Chanda et al., 2010; Schlosburg et al., 2010; Taschler et al., 2011). Tolerance to the context-dependent anxiolytic effects in rats of a lower dose of JZL184 (8 mg·kg ^{for 6 days) was not, however, seen (}Busquets-Garcia et al., 2011; Sciolino et al., 2011), suggesting that the tolerance may require sustained high levels of MGL inhibition to occur.

Probing the (patho)physiological role of MGL

The commercial availability of selective compounds like JZL184, together with the generation of MGL ^{mice, has resulted in a veritable explosion of our knowledge concerning the consequences of MGL inhibition both} in vitro and in vivo in animal models of physiological function and pathological dysfunction as diverse as colitis, pain, cancer and Parkinson's disease (see Table 1 for a comprehensive list of the in vivo effects of MGL inhibitors and of MGL genetic deletion). Rather than repeat the data given in the table, this section will focus on two areas, namely retrograde signalling in vitro and the behavioural consequences of selective MGL and FAAH inhibition versus non-selective inhibition of both enzymes in vivo.

Teasing out the eCBs involved in retrograde signalling in the brain

In the brain, eCBs act in a retrograde manner, that is, are released post-synaptically and interact with presynaptic CB₁ receptors to modulate neurotransmitter release (Alger and Kim, 2011; see Katona and Freund, 2008 for an elegant animation of retrograde signalling in glutamatergic synapses). In general, retrograde signalling can be investigated electrophysiologically in tissue slices by measurement of processes such as depolarization-induced suppression of excitation (DSE) or inhibition (DSI), with the eCB component being demonstrated by the loss of the response in the presence of a CB₁ receptor antagonist/inverse agonist such as rimonabant, AM251 and AM281 (Kreitzer and Regehr, 2001; Ohno-Shosaku et al., 2001; Wilson and Nicoll, 2001). What was not initially clear, however, was whether AEA and 2-AG played separate roles in retrograde signalling pathways in different brain regions, or whether they acted simply as alternative signalling molecules within the same system. Histochemical studies in the rat hippocampus, cerebellum and amygdala indicated that MGL was mainly presynaptically localized, whereas FAAH was post-synaptic (Gulyas et al., 2004) and a presynaptic localization of MGL in the human hippocampus has also been reported (Ludányi et al., 2011). A presynaptic localization is more logical in terms of termination of the retrograde signalling molecule, and initial evidence favouring 2-AG over AEA was the finding that the selective FAAH inhibitor URB597 (structure, see Figure 2) did not affect DSI in the hippocampus (Kim and Alger, 2004). Blockade of diacylglycerol lipase α with a low-potency compound (RHC-80267) was without effect upon DSI at a concentration of 100 µM, but it did block hippocampal long-term potentiation (Chevaleyre and Castillo, 2003), a finding consistent with data demonstrating that 2-AG is formed in hippocampal slices by high-frequency stimulation of the Schaffer collaterals (Stella et al., 1997).

The availability of MGL (and ABHD6) inhibitors, as well as the use of compounds or genetic deletion strategies to block 2-AG synthesis, have demonstrated convincingly that in rodents, 2-AG is the eCB primarily responsible for retrograde signalling contributing to DSE and DSI in the cerebellum and hippocampus (Szabo et al., 2006; Hashimotodani et al., 2007; Pan et al., 2009; Gao et al., 2010), DSI in the thalamus and prefrontal cortex (Sun et al., 2011; Yoshino et al., 2011) and long-term depression in the prefrontal cortex (Lafourcade et al., 2007; Marrs et al., 2010). This role of 2-AG is not confined to rodents (or even the brain, for that matter), and has been reported for the retrograde inhibition of calcium-activated potassium channels in goldfish retinal cones (Fan and Yazulla, 2007). MGL ^{mice have also been used to confirm 2-AG involvement in hippocampal DSI and cerebellar DSE (}Pan et al., 2011; Zhong et al., 2011), although the down-regulation of CB₁ receptors seen in these animals is a complicating factor, as is the ability of cyclooxygenase-2 to contribute to the metabolism of eCBs in preparations used in electrophysiological experiments (Kim and Alger, 2004; Slanina and Schweitzer, 2005; Fan and Yazulla, 2007; Yang et al., 2008b; Straiker and Mackie, 2009).

The predominant role of 2-AG does not mean that AEA is redundant as an eCB signalling molecule in the brain, rather that it has other functions, as has been demonstrated convincingly in many electrophysiological and behavioural experiments using FAAH inhibitors and FAAH ^{mice (e.g.} Kathuria et al., 2003; Manwell et al., 2009; Schlosburg et al., 2009; Bambico et al., 2010; Kinsey et al., 2010; Ramesh et al., 2011).

‘Cannabis-like’ effect of MGL inhibitors?

A concern with the development of molecules affecting eCB metabolism is that they might produce cannabis-like effects, which would limit their usefulness. This concern was greatly reduced when the first selective FAAH inhibitor, URB597, was found not to produce catalepsy, hypothermia or hyperphagia, responses that are seen with compounds activating central CB₁ receptors (Kathuria et al., 2003), and not to produce reinforcing effects in monkeys trained to self-administer Δ^{-THC (}Justinova et al., 2008). A similar lack of ‘cannabis-like’ behaviour has been seen with other selective FAAH inhibitors, such as PF-3845 (structure, see Figure 2; Long et al., 2009c). With respect to MGL inhibition, JZL184, at a dose of 16 mg·kg ^{i.p., was found to produce rimonabant-sensitive thermal hypoalgesia and hypolocomotion in mice, but it did not produce any motor impairments on the rotor rod and catalepsy bar tests, suggesting that at a dose producing a sustained increase in brain 2-AG levels, the full gamut of behavioural responses that characterize a general activation of brain CB}₁ receptors is not seen (Schlosburg et al., 2009; Long et al., 2009a,b). A similar pattern was seen at a higher dose (40 mg·kg ^i.p.; Long et al., 2009a), and at a dose of 8 mg·kg ^{i.p., JZL184 did not produce catalepsy in the bar test in rats following either a single or a repeated (6-day) dosage regime (}Sciolino et al., 2011).

The difference in the pattern of behavioural responses between FAAH and MGL inhibition underlines the point made above that in the brain, AEA and 2-AG have different functions rather than simply being alternate ligands for a given target. However, when both enzymes were inhibited, either by the combination of JZL184 and PF-3845, or by administration of the non-selective MGL and FAAH inhibitor JZL195 (structure, see Figure 2), dramatic effects upon all four measures of the behavioural ‘tetrad’ (thermal analgesia, catalepsy, locomotor activity and rectal temperature) were seen; effects which in all cases were prevented by rimonabant (Long et al., 2009c). Furthermore, while JZL184 produced only a partial degree of Δ^{-THC-appropriate responding in a drug discrimination assay using FAAH mice, a complete mirror of the Δ-THC response was seen with this drug in FAAH mice (}Long et al., 2009c). Although these results were obtained at high levels of sustained MGL and FAAH blockade, selectivity of inhibition for MGL versus FAAH (or vice versa) appears to be a prerequisite for drug development to avoid overt cannabis-like effects of the drugs.

Clinical potential of MGL inhibitors

Examination of Table 1 would suggest that on the basis of the hitherto undertaken preclinical pharmacology of MGL inhibitors, several different indications can be considered. When the list is restricted to those where several independent studies have supported one another, pain and cancer are the two strongest current potential indications for such compounds. The anxiolytic effects of MGL inhibitors are also of interest, although the context dependency of the effects (i.e. how strong the aversive stimulus is) is a complicating factor in assessing clinical potential. In consequence, the following section is limited to consideration of pain and cancer as potential indications for MGL inhibitors.

Cancer

Ever since the first demonstration that Δ^{-THC and other phytocannabinoids reduced the rate of growth of lung tumour xenografts (}Munson et al., 1975), the potential of cannabinoids as anticancer agents has been actively explored and mechanism(s) responsible for their antitumoural effects have been investigated in detail (reviews, see Velasco et al., 2004; Flygare and Sander, 2008; Sarfaraz et al., 2008; Freimuth et al., 2010; Fowler et al., 2010b; Díaz-Laviada, 2011; Malfitano et al., 2011). In many (but by no means all) cancer forms, increased expression of CB receptors is seen in tumours, and in prostate tumours, this increase may have prognostic usefulness in aiding treatment decisions (Fowler et al., 2010a; see Figure 3). It is not clear whether a changed tumour CB₁ expression is a result of malignancy or alternatively a potentially damaging selection factor (by contributing to the ability of the tumour to survive in its microenvironment, see Cudaback et al., 2010). Nonetheless, it is indicative of a disturbed local eCB signalling, and thus consideration of the potential of MGL inhibition in cancer is motivated, particularly in the light of the finding that VDM11 (N-(4-hydroxy-2-methylphenyl) arachidonoyl amide), which potentiates eCBs by preventing their cellular uptake, reduces the growth of thyroid cancer cells in vivo in a xenograft model (Bifulco et al., 2004).

Figure 3

CB₁ receptor expression in prostate cancer. (A) An example of strong CB₁ receptor immunoreactivity (CB₁IR) in the epithelia, but not stroma of a formalin-fixed, paraffin-embedded sample of prostate cancer tissue obtained at transurethral resection for voiding difficulties, when the diagnosis was made (photograph of A. Bergh, S.C. Chung, C.J. Fowler; dataset published in Chung et al., 2009). (B) Influence of CB₁IR with other parameters on disease-specific survival of patients (n= 234) who were followed by watchful waiting after diagnosis until the appearance of metastases. The Exp(B) values, determined by multivariate Cox proportional-hazards regression analysis, refers to the odds ratio, that is the change in odds as a result of an increase in the ‘unit’ of the predictive variable under study, with a positive value as here indicative of a deleterious influence on disease outcome. For CB₁ receptors, the value Exp(B) is 2.64 (95% confidence limits 1.53–4.58) for the 36 cases with a CB₁IR score above the optimum cut-off value compared to the 198 cases with a score below this cut-off. pEGFR refers to phosphorylated epidermal growth factor receptor expression, T2-T4 the tumour stage and GS the Gleason scores. Significance levels: *P < 0.05, **P < 0.01. Data from table 1 of Fowler et al. (2010a).

In an elegant study from 2004, the effect of changed endogenous eCB concentrations upon the ability of prostate cancer cell lines was investigated (Nithipatikom et al., 2004). They found that the CB receptor agonist noladin ether reduced the in vitro invasivity of androgen-independent human DU145 and PC-3, but not androgen-sensitive LNCaP prostate cancer cells. More importantly in the present context, the authors reported that the 2-AG synthesis inhibitor RHC-80267, at a concentration which greatly reduced 2-AG levels, increased the invasivity of PC-3 and DU-145, but not LNCaP cells in the in vitro model (Nithipatikom et al., 2004). The converse was seen for PC-3 cells treated with MAFP, with other compounds found to inhibit 2-AG hydrolysis (including JZL184), and following knockdown of MGL (Nithipatikom et al., 2004; 2005; Nomura et al., 2011a), suggesting that there is an eCB tonus controlling PC-3 invasivity that can be potentiated. Conversely, rimonabant increased the invasive behaviour of PC-3 and DU-145, but not LNCaP cells in vitro (Nithipatikom et al., 2004), but it is not clear whether this effect is related to the blockade of an eCB tonus or the inverse agonist properties of this compound. In a xenograft model using PC-3 cells injected into the flanks of immunodeficient mice, reduction of MGL activity either using JZL184 or genetic knockdown of the tumour cell enzyme resulted in a slowed tumour growth that in the latter case was partially blocked by rimonabant (Nomura et al., 2011a; see Table 1).

JZL184 treatment and genetic knockdown of MGL have also been found to reduce the tumour sizes in xenograft models of ovarian, melanoma and colorectal cancers (Nomura et al., 2010; Ye et al., 2011; see Table 1). The potential of MGL inhibitors in other ovarian cancer and melanoma, however, may be related to their ability to reduce the formation of long-chain fatty acids from their corresponding glycerol esters, given the wide substrate specificity of MGL. Thus, (i) the increased MGL activity seen in primary high-grade ovarian tumours compared to benign tumours was accompanied by higher levels of long-chain free fatty acids; (ii) reduction of MGL activity in melanoma cancer cell lines decreased long-chain free fatty acid levels, whereas the converse was seen when MGL was transfected into melanoma cells. The latter also increased tumour growth in a xenograft model; (iii) reduced migration of ovarian and melanoma tumours in vitro was produced by reduced MGL activity in a manner overridden by addition of a long-chain free fatty acid; (iv) reduced tumour growth in a xenograft model was seen following knockdown of MGL in the melanoma cells used. The reduced tumour growth was overridden by a high-fat diet (Nomura et al., 2010). Even in the prostate cancer cells, long-chain free fatty acids are important, given that the high-fat diet also partially overrides the effect of tumour MGL disruption upon xenograft tumour growth, with a complete reversal being seen with the combination of rimonabant and a high-fat diet (Nomura et al., 2011a). Although the predictive value of xenografts should not be overestimated (Kerbel, 2003), these results are encouraging, and a non-eCB-mediated mechanism of antitumoural action of MGL inhibitors would reduce the potential importance of tolerance described earlier, the assumption of course being that a down-regulation of eCB signalling due to tolerance is not detrimental in itself.

Pain

Given that MGL inhibitors represent a novel therapeutic approach, the choice of the initial phase II clinical trial will be a ‘best guess’ as to which patient group should be targeted. This is critical, given that the outcome of the first trial is likely to affect not only development of the compound in question, but also the willingness of other pharmaceutical companies to invest in MGL inhibitors. In this author's view, three criteria should be fulfilled in order to optimize the ‘best guess’:

• the pain state(s) under study should have been shown in clinical trials to respond to cannabinoids. This at least indicates that the indication in question is tractable to modulation of eCB signalling. It should be noted that in this respect, cannabinoids refer to ‘pure’ cannabinoid preparations or drugs such as nabilone, rather than extracts like Sativex, since in the extracts, the presence of cannabidiol may in theory contribute to the observed clinical effects.

• experimental studies should indicate that the eCB system is out of balance in animal models of the disease in question (or preferably in the disease itself), so that the MGL inhibitor has a measurable effect, either to normalize a deficient signal, or to potentiate a protective response. MGL inhibitors might of course work if the system is in balance, but the lack of effect of Δ^{-THC in experimental clinical pain (see below) does not make this author optimistic in this regard.}

• for indications requiring repeat dosage regimes, there should be no tolerance to the compound in question.

It can be noted that ‘positive effects of the MGL inhibitor in animal models of the indication in question’ is not among my criteria. This criterion has been omitted for two reasons: (i) it is extremely unlikely that a clinical trial would be considered in the absence of such data; thus the criterion is redundant; (ii) the criterion assumes that the preclinical data is entirely predictive of the situation in the clinic, which would render the first two of my criteria as irrelevant. However, the preclinical promise of novel analgesic drugs has not resulted in very much clinical success, particularly in the case of drugs directed towards neuropathic pain. Although this is due in part to the heterogeneity of the clinical population targeted in clinical trials (Woolf, 2010), there is a mismatch between the outcome measures used in standard animal models and those used in clinical trials (Rice et al., 2008). Hopefully, the development of animal models which mirror more closely the clinical situation and the use of more homogenous patient populations to target pain indications relevant to the mechanism of action of the drug in question (Woolf, 2010) will remedy this situation.

The potential pain indications for MGL inhibitors discussed below will be considered with respect to these criteria.

Single-/limited-dose pain indications for MGL inhibitors

Single-dose studies are sometimes used in experimental settings as a short cut to seeing if the compound in question is active in man. In the pain field, there have been several studies investigating Δ^{-THC in experimental pain settings (such as warm and cold skin stimulation, cold pressor test, intradermal capsaicin-inducing pain). The studies have been well reviewed elsewhere (}Staahl et al., 2009), where the authors concluded that ‘The analgesic effect of delta-9- tetrahydrocannabinol is hard to show in acute experimental pain models’, and further that the effects as such could be upon sensory/discriminative rather than affective/motivational components of pain. This should be taken into account if experimental human pain is used to assess the potential of MGL inhibitors.

With respect to alleviation of post-operative pain, the available data does not point to a robust effect of cannabinoids in postoperative pain (indeed, in one case, a hyperalgesic response to a high dose of the cannabinoid nabilone was reported; Beaulieu, 2006). There are issues of sample size, dosing techniques and not least the time after surgery when the cannabinoid was given, which are limiting factors in the interpretation of these clinical trials (for review, see Beaulieu and Ware, 2007). Nonetheless, current data would argue against the use of post-operative pain as an indication with which to test for clinical efficacy of MGL inhibitors, although a more complex study of the effect of MGL inhibition upon the development of persistent pain following surgery may be warranted (see Box 3).

Persistent pain indications for MGL inhibitors

Cannabinoids were given a rather poor write-up in 2001, when they were described as ‘no more effective than codeine in controlling pain and have depressant effects on the central nervous system that limit their use’ (Campbell et al., 2001). However, since then, several high-quality randomized clinical trials have been conducted, and these have led the first author of Campbell et al. (2001) instead to conclude that ‘Overall there is evidence that cannabinoids are safe and modestly effective in neuropathic pain with preliminary evidence of efficacy in fibromyalgia and rheumatoid arthritis’ (Lynch and Campbell, 2011). Most of these studies used cannabis extracts, that is assessed the effect of CB receptor activation in the presence of cannabidiol, but there were positive outcomes in five placebo-controlled studies using nabilone or dronabinol in patients with chronic pain (Lynch and Campbell, 2011). Cancer pain was not considered in this article, but there are reports of analgesic efficacy of nabilone in advanced cancer patients (Maida et al., 2008) and of a Δ^{-THC : cannabidiol extract (but not of Δ-THC alone at the dosing used) in patients with intractable cancer-related pain (}Johnson et al., 2010).

From the above, several persistent pain states satisfy the first criterion in that they are responsive to cannabinoids. Less, however, is known about the eCB system itself in human pain states. However, detectable levels of AEA and 2-AG were found in synovial fluid from patients with osteoarthritis and rheumatoid arthritis undergoing total knee arthroplasty, whereas no detectable levels were seen in synovial fluid from normal volunteers (Richardson et al., 2008). The eCBs were suggested to come from immune cells present in the synovial fluid rather than being a reflection of eCB release and turnover in the synovium (Richardson et al., 2008). Increased CB₁ receptor immunoreactivity has also been reported in ventral tendon biopsies from patients with Achilles tendinosis compared to control biopsies² (Björklund et al., 2011), and an association between pain scores and tumour CB₁ receptor expression (but not MGL expression) has been reported for patients with pancreatic cancer (Michalski et al., 2008). Whether or not these differences are adaptive changes, responses aimed at mitigating the pain, or alternatively changes that contribute to the disorder requires elucidation, but at least they suggest that the eCB system is changed in pain states, thereby giving some tenuous support to the second criterion. A word of caution, however: potential effects produced by blocking 2-AG metabolism may be mitigated if the 2-AG finds alternative metabolic routes. The obvious example is cyclooxygenase-2 (Duggan et al., 2011), particularly since the metabolic product, prostaglandin glycerol ester, can produce hyperalgesia when given exogenously to experimental animals (Hu et al., 2008)³. Currently, data suggesting that cyclooxygenase-2 contributes significantly to eCB metabolism has been from in vitro experiments, although there is in vivo data suggesting that nimesulide can affect the amount of AEA available for metabolism by FAAH in the mouse brain (Glaser and Kaczocha, 2010).

The third criterion, that of tolerance, may be problematic given the data with JZL184 and with MGL ^{mice (}Chanda et al., 2010; Schlosburg et al., 2010; Taschler et al., 2011). However, in these cases, a complete and long-lasting block of MGL produced by the genetic deletion or irreversible inhibition of the enzyme may simply be a case of raising 2-AG levels for too long and too excessively. Certainly, the lack of tolerance to the context-dependent anxiolytic effects of JZL184 in rats and mice and the antinociceptive effects of the compound in a mouse visceral pain model with a lower dose (8 mg·kg^{) of the compound (}Busquets-Garcia et al., 2011; Sciolino et al., 2011) are consistent with this conclusion. The field clearly needs to know whether reversible inhibitors (with appropriate pharmacokinetic properties, to avoid long-lasting block of MGL) also produce dose-dependent tolerance, before an informed decision upon whether or not pain is a suitable indication for MGL inhibitors can be made.

The discussion above has considered MGL inhibitors as a stand-alone treatment for pain. However, MGL inhibitors may also be useful in conjunction with non-steroidal anti-inflammatory drugs (NSAIDs), since JZL184 (4 mg·kg^{) provides protection against gastric haemorrhages produced by diclofenac in mice. This protection, which is not seen in the absence of CB}₁ receptors, persists (albeit to a slightly lesser extent) even when the animals were pretreated with the MGL inhibitor for 5 days prior to diclofenac treatment (Kinsey et al., 2010). The FAAH inhibitor URB597 shows the same property and also acts synergistically with diclofenac in a model of visceral pain (Naidu et al., 2009). It is not known whether such synergy is also seen with MGL inhibitors, but strategies resulting in lower doses of NSAIDs and with added gastroprotection would clearly be of benefit in the clinic.

Cancer

Ever since the first demonstration that Δ^{-THC and other phytocannabinoids reduced the rate of growth of lung tumour xenografts (}Munson et al., 1975), the potential of cannabinoids as anticancer agents has been actively explored and mechanism(s) responsible for their antitumoural effects have been investigated in detail (reviews, see Velasco et al., 2004; Flygare and Sander, 2008; Sarfaraz et al., 2008; Freimuth et al., 2010; Fowler et al., 2010b; Díaz-Laviada, 2011; Malfitano et al., 2011). In many (but by no means all) cancer forms, increased expression of CB receptors is seen in tumours, and in prostate tumours, this increase may have prognostic usefulness in aiding treatment decisions (Fowler et al., 2010a; see Figure 3). It is not clear whether a changed tumour CB₁ expression is a result of malignancy or alternatively a potentially damaging selection factor (by contributing to the ability of the tumour to survive in its microenvironment, see Cudaback et al., 2010). Nonetheless, it is indicative of a disturbed local eCB signalling, and thus consideration of the potential of MGL inhibition in cancer is motivated, particularly in the light of the finding that VDM11 (N-(4-hydroxy-2-methylphenyl) arachidonoyl amide), which potentiates eCBs by preventing their cellular uptake, reduces the growth of thyroid cancer cells in vivo in a xenograft model (Bifulco et al., 2004).

Figure 3

CB₁ receptor expression in prostate cancer. (A) An example of strong CB₁ receptor immunoreactivity (CB₁IR) in the epithelia, but not stroma of a formalin-fixed, paraffin-embedded sample of prostate cancer tissue obtained at transurethral resection for voiding difficulties, when the diagnosis was made (photograph of A. Bergh, S.C. Chung, C.J. Fowler; dataset published in Chung et al., 2009). (B) Influence of CB₁IR with other parameters on disease-specific survival of patients (n= 234) who were followed by watchful waiting after diagnosis until the appearance of metastases. The Exp(B) values, determined by multivariate Cox proportional-hazards regression analysis, refers to the odds ratio, that is the change in odds as a result of an increase in the ‘unit’ of the predictive variable under study, with a positive value as here indicative of a deleterious influence on disease outcome. For CB₁ receptors, the value Exp(B) is 2.64 (95% confidence limits 1.53–4.58) for the 36 cases with a CB₁IR score above the optimum cut-off value compared to the 198 cases with a score below this cut-off. pEGFR refers to phosphorylated epidermal growth factor receptor expression, T2-T4 the tumour stage and GS the Gleason scores. Significance levels: *P < 0.05, **P < 0.01. Data from table 1 of Fowler et al. (2010a).

In an elegant study from 2004, the effect of changed endogenous eCB concentrations upon the ability of prostate cancer cell lines was investigated (Nithipatikom et al., 2004). They found that the CB receptor agonist noladin ether reduced the in vitro invasivity of androgen-independent human DU145 and PC-3, but not androgen-sensitive LNCaP prostate cancer cells. More importantly in the present context, the authors reported that the 2-AG synthesis inhibitor RHC-80267, at a concentration which greatly reduced 2-AG levels, increased the invasivity of PC-3 and DU-145, but not LNCaP cells in the in vitro model (Nithipatikom et al., 2004). The converse was seen for PC-3 cells treated with MAFP, with other compounds found to inhibit 2-AG hydrolysis (including JZL184), and following knockdown of MGL (Nithipatikom et al., 2004; 2005; Nomura et al., 2011a), suggesting that there is an eCB tonus controlling PC-3 invasivity that can be potentiated. Conversely, rimonabant increased the invasive behaviour of PC-3 and DU-145, but not LNCaP cells in vitro (Nithipatikom et al., 2004), but it is not clear whether this effect is related to the blockade of an eCB tonus or the inverse agonist properties of this compound. In a xenograft model using PC-3 cells injected into the flanks of immunodeficient mice, reduction of MGL activity either using JZL184 or genetic knockdown of the tumour cell enzyme resulted in a slowed tumour growth that in the latter case was partially blocked by rimonabant (Nomura et al., 2011a; see Table 1).

JZL184 treatment and genetic knockdown of MGL have also been found to reduce the tumour sizes in xenograft models of ovarian, melanoma and colorectal cancers (Nomura et al., 2010; Ye et al., 2011; see Table 1). The potential of MGL inhibitors in other ovarian cancer and melanoma, however, may be related to their ability to reduce the formation of long-chain fatty acids from their corresponding glycerol esters, given the wide substrate specificity of MGL. Thus, (i) the increased MGL activity seen in primary high-grade ovarian tumours compared to benign tumours was accompanied by higher levels of long-chain free fatty acids; (ii) reduction of MGL activity in melanoma cancer cell lines decreased long-chain free fatty acid levels, whereas the converse was seen when MGL was transfected into melanoma cells. The latter also increased tumour growth in a xenograft model; (iii) reduced migration of ovarian and melanoma tumours in vitro was produced by reduced MGL activity in a manner overridden by addition of a long-chain free fatty acid; (iv) reduced tumour growth in a xenograft model was seen following knockdown of MGL in the melanoma cells used. The reduced tumour growth was overridden by a high-fat diet (Nomura et al., 2010). Even in the prostate cancer cells, long-chain free fatty acids are important, given that the high-fat diet also partially overrides the effect of tumour MGL disruption upon xenograft tumour growth, with a complete reversal being seen with the combination of rimonabant and a high-fat diet (Nomura et al., 2011a). Although the predictive value of xenografts should not be overestimated (Kerbel, 2003), these results are encouraging, and a non-eCB-mediated mechanism of antitumoural action of MGL inhibitors would reduce the potential importance of tolerance described earlier, the assumption of course being that a down-regulation of eCB signalling due to tolerance is not detrimental in itself.

Conclusions

In just a few years, the availability of selective MGL inhibitors has allowed for the identification of the roles played by 2-AG in the body, and provided preclinical data supporting potential indications for this class of compounds. Within the pain field, issues of tolerance and ‘cannabis-like’ effects secondary to loss of selectivity vis á vis FAAH following repeated high dosing may present obstacles to drug development, but hopefully in vivo data with reversible inhibitors will be forthcoming to address these issues. Within the cancer field, it is early days and in vivo data with genetic and orthotopic cancer models would be most useful, as would combination studies with standard drug regimes. A separate issue is whether potentiation of eCBs following MGL inhibition may produce unwanted effects upon immune function, given that CB₂ receptor activation can dramatically affect the properties of immune cells (Basu and Dittel, 2011). Nonetheless, when it is remembered that prior to 2005, no selective and ‘drugable’ MGL inhibitors were available, the rapid progress in the field is impressive, and gives hope that the therapeutic potential of MGL as a target for drug development will be translated into a clinical reality.