Druggable Targets of the Endocannabinoid System: Implications for the Treatment of HIV-Associated Neurocognitive Disorder

Abstract

HIV-associated neurocognitive disorder (HAND) affects nearly half of all HIV-infected individuals. Synaptodendritic damage correlates with neurocognitive decline in HAND, and many studies have demonstrated that HIV-induced neuronal injury results from excitotoxic and inflammatory mechanisms. The endocannabinoid (eCB) system provides on-demand protection against excitotoxicity and neuroinflammation. Here, we discuss evidence of the neuroprotective and anti-inflammatory properties of the eCB system from in vitro and in vivo studies. We examine the pharmacology of the eCB system and evaluate the therapeutic potential of drugs that modulate eCB signaling to treat HAND. Finally, we provide perspective on the need for additional studies to clarify the role of the eCB system in HIV neurotoxicity and speculate that strategies that enhance eCB signaling might slow cognitive decline in HAND.

1. Introduction

Nearly half of all HIV-infected individuals experience cognitive and motor impairments (Heaton et al., 1995; Tozzi et al., 2005); these symptoms are collectively termed HIV-associated neurocognitive disorder (HAND). Combined antiretroviral therapy (cART) has reduced the number of patients that progress to AIDS (Heaton et al., 2010; McArthur, 1993), and increased the lifespan of infected individuals (ATCC, 2008; Dore et al., 2003). However, while cART suppresses viral load, it does not eradicate HIV from the brain (Valcour et al., 2011). Thus, the prevalence of HAND remains high (Cysique et al., 2004; Sacktor et al., 2001; Saylor et al., 2016), with no effective treatment available.

One promising area of therapeutic development is the endocannabinoid (eCB) system. This biological system is composed of endogenous lipid-based signaling molecules that bind to cannabinoid receptors, and the machinery that synthesizes and metabolizes them (Fig. 1). eCB signaling is dependent on neurons interacting with other neurons, astrocytes, and microglia (Ashton and Glass, 2007; Navarrete et al., 2014). The eCB system plays a role in many physiological processes. Here, we review its ability to provide on-demand protection against excitotoxicity and neuroinflammation (Marsicano et al., 2003; Walter and Stella, 2004), two hallmarks of many neurodegenerative disorders including HAND (Amor et al., 2010; Dong et al., 2009; Green et al., 2018).

Fig. 1:

Neuroprotective targets in the eCB system. Schematic displays principal neuron receiving excitatory and inhibitory input with neighboring astrocyte and microglial cells. Cannabinoid type 1 receptors (CB₁Rs) are present on presynaptic terminals, with high expression on a subset of GABAergic terminals and widespread expression at lower levels on glutamatergic terminals (Marsicano and Lutz, 1999). Cannabinoid type 2 receptors (CB₂Rs) are found predominantly on cells of the immune system (microglia shown), with limited expression on neurons (not shown) (Nunez et al., 2004). All four cell types express diacylglycerol lipase (DGL) and N-arachidonoyl phosphatidylethanolamine phospholipase D (NAPE-PLD), the enzymes that synthesize the eCBs 2-arachdionoylglycerol (2-AG) and anandamide (AEA), respectively (Egertova et al., 2008; Ludanyi et al., 2011; Mishra et al., 2016; Viader et al., 2016; Zhang et al., 2011). For clarity, the pathways are described in detail in a single location with high expression. 2-AG is hydrolyzed by monoacylglycerol lipase (MGL) that is expressed in all four cell types, with the highest levels found in astrocytes (Muccioli et al., 2007; Viader et al., 2015). Fatty acid amide hydrolase (FAAH) is predominantly expressed in the somata and dendrites of principal neurons (Egertova et al., 2003; Tsou et al., 1998b). Important neuroprotective drug targets (blue boxes) include agonists for CB₁ and CB₂Rs (dashed green lines) and inhibitors of MGL and FAAH (dashed red lines). Abbreviations; 2-AG: 2-arachidonoylglycerol; AMPAR: α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor; AA: arachidonic acid; AEA: arachidonoyl ethanolamine (anandamide); CB₁R: cannabinoid type 1 receptor; CB₂R: cannabinoid type 2 receptor; DAG: diacylglycerol; DGL: diacylglycerol lipase; FAAH: fatty acid amide hydrolase; GABA: γ-aminobutyric acid; GABA_AR: GABA type A receptor; IP₃: inositol triphosphate; mGluR: metabotropic glutamate receptor; MGL: monoacylglycerol lipase; NAPE-PLD: N-arachidonoyl phosphatidylethanolamine phospholipase D; NAT: N-acyltransferase; NAPE: N-arachidonoyl phosphatidylethanolamine; PE: phosphatidylethanolamine; PG: prostaglandin; PIP₂: phosphatidylinositol 4,5-bisphosphate; PLC: phospholipase C; VGCC: voltage-gated Ca²⁺ channel.

Drugs can enhance the eCB system by mimicking or potentiating its neuroprotective function. Changes in the eCB system associated with some neurological disorders might impair the neuroprotection that the system affords, exacerbating excitotoxicity. Thus, strategies to prevent loss of eCB signaling may preserve its neuroprotective function, and drugs that mimic or enhance eCB signaling may compensate for its diminished capacity in neurotoxic or inflammatory conditions, such as HAND. Cannabimimetic drugs are relevant to HIV because they are used to treat wasting and nausea in AIDS patients (Plasse et al., 1991), cannabis use among HIV-positive individuals is high (Okafor et al., 2017), and recent efforts to legalize medicinal and recreational marijuana are increasing access (Cerda et al., 2017). Here, we examine the components and pharmacology of the eCB system and discuss evidence of its neuroprotective and anti-inflammatory properties. We then evaluate the therapeutic potential of drugs that modulate the eCB system and speculate that strategies that enhance eCB signaling might slow cognitive decline in HAND.

2. Cannabinoid Receptors

CB₁ and CB₂ receptors have been studied extensively and are the best understood receptor components of the eCB system. However, eCBs also interact with ion channels and other G protein-coupled receptors (GPCRs). Anandamide activates the transient receptor potential vanilloid 1 (TRPV1) receptor, a calcium permeable, nonselective cation channel (Ross, 2003). Several orphan GPCRs, notably GPR55 and GPR18, have emerged as possible targets of eCBs (for review, see (Cuevas-Olguin et al., 2017)). While eCBs may act on these alternative targets in a manner relevant to pathophysiology, this review will focus on CB₁ and CB₂ receptors.

2.1. CB₁ receptor agonists

Cannabinoid type 1 receptors (CB₁Rs) are GPCRs abundantly expressed on neurons and astrocytes (Howlett et al., 2002; Matsuda et al., 1993; Navarrete and Araque, 2008) (Fig. 1). CB₁Rs on neurons localize to presynaptic elements, where they modulate neurotransmission. CB₁Rs on astrocytes modulate gliotransmission by regulating intracellular calcium, facilitating communication with neurons (Navarrete and Araque, 2008; Navarrete and Araque, 2010; Oliveira da Cruz et al., 2016). Histochemical studies have found CB₁R mRNA in the principal (excitatory) neurons of the cerebellum, hypothalamus, thalamus, and lower brain stem (Howlett et al., 2002; Tsou et al., 1998a). CB₁Rs are also found in a subset of inhibitory neurons, particularly cholecystokinin (CCK)-positive interneurons of the hippocampus, amygdala, and cerebral cortex (Herkenham et al., 1991; Mackie, 2005), with no expression in parvalbumin-positive interneurons (Marsicano and Lutz, 1999; Matsuda et al., 1993). This distribution corresponds to the physiological role of eCBs in control of motor activity, nociception, learning and memory, food intake, and cognitive function (for review, see (Di Marzo et al., 1998)). Most CB₁Rs are coupled to G_i/o (Pertwee et al., 2010), where activation of the G_αi/G_αo subunit reduces cAMP production by inhibiting adenylyl cyclase, while the G_βγ complex targets many physiological effectors, including inwardly rectifying K⁺ channels, N- and P/Q-type Ca²⁺ channels, protein kinase signaling cascades, and others (Kano et al., 2009; Smrcka, 2008).

Here, we focus on the neuroprotective role of CB₁Rs that, upon activation, inhibit glutamate release at excitatory synapses (Gerdeman and Lovinger, 2001; Shen et al., 1996). These receptors are part of a retrograde signaling system in which eCBs produced postsynaptically diffuse across the synaptic cleft to act on presynaptic CB₁Rs. At glutamatergic synapses, this system provides a form of feedback inhibition. Upon excitation, the postsynaptic cell synthesizes the eCB 2-arachidonoylglycerol (2-AG) that then activates presynaptic CB₁Rs to inhibit and reduce glutamate release. Interestingly, a recent study showed that astrocytes produce 2-AG in response to activation of metabotropic glutamate receptors, with subsequent activation of presynaptic CB₁R to produce a form of synaptic depression (Smith et al., 2019).

CB₁R-mediated presynaptic inhibition provides a basis for the neuroprotective properties of the eCB system, whereby the ability to inhibit glutamate release dampens excitotoxicity caused by excessive activation of glutamatergic pathways. Excitotoxicity is a hallmark of many neurodegenerative disorders (Dong et al., 2009) including HAND (Green et al., 2018), where overstimulation of glutamate receptors leads to a characteristic loss of postsynaptic structures (Dong et al., 2009); thus, preventing it may slow disease progression and attenuate symptoms.

Compelling evidence that CB₁Rs protect against excitotoxicity comes from Marsicano et al. (2003), who demonstrated that mice lacking CB₁Rs in all principal neurons of the forebrain but not astrocytes experienced more severe seizures in a kainic acid model of excitotoxic epileptiform seizures (Marsicano et al., 2003). Furthermore, Monory et al. (2006) demonstrated that CB₁Rs on hippocampal glutamatergic neurons were central mediators of on-demand eCB-dependent protection against kainic acid-induced acute excitotoxic seizures (Monory et al., 2006). Subsequent reports expanded upon these results, demonstrating that viral-mediated overexpression of CB₁Rs in the principal neurons of the hippocampus protected against seizure-induced excitotoxicity (Guggenhuber et al., 2010), and that the eCB 2-AG is the key activator of the CB₁Rs that ultimately mediate seizure suppression (Sugaya et al., 2016).

In addition to preventing excitotoxicity by suppressing aberrant patterns of glutamatergic activity, CB₁R-mediated neuroprotection may also result from activation of cell survival signaling cascades, including the phosphatidylinositol 3-kinase (PI3K)/Akt pathway and the extracellular signal-regulated (ERK) pathway (Dalton and Howlett, 2012; Molina-Holgado et al., 2002). Thus, CB₁R agonists may be viable therapeutic options to attenuate changes in network function that contribute to neuronal damage. Indeed, CB₁R agonists attenuate excitotoxicity in several models of neurodegenerative disorders. The cannabinoid receptor agonist WIN55,212–2 reduced synaptically-mediated excitotoxicity evoked by reduction of ambient [Mg²⁺] in primary hippocampal cultures; this effect was blocked by the CB₁R inverse agonist rimonabant (SR141716A), indicating a role for CB₁Rs (Shen and Thayer, 1998). In rodent models of traumatic brain injury and stroke, pre-treatment with the cannabinoid agonists WIN55,212–2 and 2-AG reduced brain edema and infarct volumes (Nagayama et al., 1999; Panikashvili et al., 2001), and induced faster recovery of motor and behavioral function (Panikashvili et al., 2001). CB₁R activation slowed disease progression in a model of multiple sclerosis (de Lago et al., 2012) possibly via suppression of inflammation, although excitotoxic mechanisms also participate in this complex model. In both in vitro and in vivo models of Parkinson’s and Huntington’s diseases, administration of exogenous 2-AG attenuated 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)- and mutant huntingtin-induced neurodegeneration, respectively (Mounsey et al., 2015; Scotter et al., 2010). Several studies have also demonstrated that cannabinoid agonists protect against amyloid β-induced alterations in neuronal function and cognitive impairment (for review, see (Aso and Ferrer, 2014)). Notably, Haghani et al. (2012) demonstrated that administration of 2-AG protected against amyloid β-induced changes in intrinsic excitability of CA1 pyramidal neurons and improved retention and recall in a passive avoidance test in mice that were bilaterally injected with the Aβ_(1–42) peptide fragment. These effects were blocked by the CB₁R antagonist AM251, but not the CB₂R antagonist AM630 (Haghani et al., 2012). In the AβPP/PS1 mouse model of Alzheimer’s disease, Aso et al. (2012) demonstrated that chronic administration of the synthetic CB₁R agonist arachidonoyl-2’-chloroethylamide (ACEA) protected neurons against Aβ toxicity and improved performance in the two-object recognition test (Aso et al., 2012). Despite these and other promising preclinical studies, results from several clinical trials on the effect of exogenous cannabinoids in relieving dementia and related symptoms were inconclusive (Krishnan et al., 2009).

The protection afforded by CB₁R agonists in neurodegeneration models suggests that they may also afford protection in HAND. Indeed, activation of CB₁Rs modulates changes in network excitability induced by the HIV protein Tat. Using an in vitro model of HIV neurotoxicity, Xu et al. (2017) demonstrated that the eCBs anandamide and 2-AG reduced Tat-induced increases in intracellular [Ca²⁺] and promoted neuronal survival; the CB₁R inverse agonist rimonabant prevented this eCB-mediated neuroprotection (Xu et al., 2017). While the mechanism of this CB₁R-mediated neuroprotection is unknown, this and other studies (discussed in subsequent sections) suggest that enhancing eCB signaling may be beneficial in HAND.

Several important considerations should be noted in regard to targeting CB₁Rs to combat HAND and other neurodegenerative disorders. Because CB₁R activation largely mediates the psychoactive properties of exogenous cannabinoids (for review, see (Howlett et al., 2002)), the therapeutic potential of CB₁R agonists is limited by adverse effects, including changes in feeding behavior (Haney et al., 2005; Wiley et al., 2005), motor slowing (More and Choi, 2015; Prashad and Filbey, 2017), and sedation (Zhornitsky and Potvin, 2012). Cannabinoid-induced cognitive impairment, particularly inhibition of short-term memory (Ranganathan and D’Souza, 2006), has also been demonstrated; this is an important consideration when the therapeutic goal is to prevent cognitive decline. Furthermore, chronic cannabis use can also produce dependence (Lopez-Quintero et al., 2011; Volkow et al., 2014). Thus, while CB₁R agonists are neuroprotective in vitro and reduce neuronal damage in animal models of neurodegenerative disease, their shortcomings have spurred interest in other components of the eCB system. Selectively modulating CB₂R and slowing eCB metabolism are two promising approaches with reduced psychoactive side effects.

2.2. CB₂ receptor agonists

Cannabinoid type 2 receptors (CB₂Rs) are G_i/o protein-coupled receptors found mainly in immune cells, including microglia, the resident macrophages of the CNS (Howlett et al., 2002; Nunez et al., 2004) (Fig. 1). Although some reports have described limited CB₂R expression in neurons (Gong et al., 2006), activation of CB₂Rs is generally thought to affect the immune system without the strong psychoactive effects of CB₁R activation. Similar to CB₁, CB₂Rs regulate signaling cascades that include adenylyl cyclase, MAP kinase, and Ca²⁺ signaling (Bayewitch et al., 1995; Slipetz et al., 1995). These pathways affect a variety of immune functions, including cytokine release (Cabral and Griffin-Thomas, 2009), cell proliferation (Carrier et al., 2004), migration (Walter et al., 2003), and gene expression (Mecha et al., 2015). CB₂Rs are undetectable in unstimulated microglia, and their expression increases in response to inflammatory stimuli (Cabral and Griffin-Thomas, 2009; Nunez et al., 2004); thus, CB₂R agonists are preferentially effective during inflammatory states. Because receptor expression is immune cell-specific and state-dependent, CB₂Rs can dampen the harmful effects of immune activation in the CNS with few adverse side effects.

Chronic inflammation is implicated in a growing number of neurodegenerative diseases. For example, reactive oxygen species produced by microglia damage neurons in Alzheimer’s and Parkinson’s disease (Javed et al., 2016; Shimohama et al., 2000). Pro-inflammatory cytokines can also damage neurons directly via excitotoxicity (Ye et al., 2013) and synaptodendritic damage (Festa et al., 2015), or indirectly through the recruitment and activation of immune cells (Kozela et al., 2011). In several disease models, CB₂R activation counteracts neuroinflammation. Neuroprotection occurs by reducing pro-inflammatory cytokines (Klegeris et al., 2003; Malek et al., 2015) and reactive oxygen species (Javed et al., 2016; Shimohama et al., 2000), increasing anti-inflammatory cytokines (Ehrhart et al., 2005; Malek et al., 2015; Molina-Holgado et al., 2003), inhibiting chemotaxis to a variety of chemokines (Romero-Sandoval et al., 2009), and shifting microglia from a pro-inflammatory M1 phenotype to an alternatively activated reparative M2a phenotype (Mecha et al., 2015). This broad range of effects is well-equipped to combat the main sources of damage during neuroinflammation. Indeed, CB₂R agonists reduce neurotoxicity and behavioral impairments in both in vitro and in vivo models of Alzheimer’s disease (Jayant et al., 2016; Ramirez et al., 2005), Parkinson’s disease (Ternianov et al., 2012), Huntington’s disease (Palazuelos et al., 2009), multiple sclerosis (Shao et al., 2014), and amyotrophic lateral sclerosis (Shoemaker et al., 2007). Thus, CB₂R agonists hold the potential to treat many neurodegenerative diseases by targeting common neuroinflammatory mechanisms.

HIV causes potentially harmful neuroinflammation (Alonso et al., 1997; Everall et al., 2009; Ubaida-Mohien et al., 2017; Walsh et al., 2014), and CB₂Rs are upregulated during HIV and simian immunodeficiency virus (SIV) infection (Benito et al., 2005; Cosenza-Nashat et al., 2011). This makes CB₂Rs an attractive target for combating HAND. Indeed, activation of CB₂Rs protects against HIV-induced neuroinflammation in vitro, mainly by decreasing inflammatory signaling and suppressing the chemokine-like activity of viral proteins (Kim et al., 2011; Purohit et al., 2014). Kim et al. (2011) demonstrated that the HIV envelope glycoprotein gp120 induced synapse loss in primary hippocampal neuronal/glial cultures. Activating CB₂Rs prevented this synapse loss by decreasing production of the pro-inflammatory cytokine IL-1β (Kim et al., 2011; Zhang and Thayer, 2018). Similarly, a higher concentration of gp120 killed neurons in striatal neuronal/glial cultures, and the cannabinoid receptor agonist WIN55,212–2 blocked production of reactive oxygen species and pro-inflammatory cytokines, reduced microglial migration, and increased neuronal survival; these effects were largely CB₂R-dependent (Hu et al., 2013).

In addition to preventing the release of neurotoxic cytokines and reactive oxygen species, CB₂R activation may actually reduce HIV infection in the CNS. Δ⁹-tetrahydrocannabinol (Δ⁹-THC) acts through CB₂Rs to reduce expression of the CD4, CXCR4, and CCR5 receptors that are required for HIV infection in monocyte-derived macrophages (Williams et al., 2014). Additionally, CB₂R activation prevents productive infection of T-tropic HIV in human CD4+ T-cells by blocking the cytoskeletal rearrangements required for viral fusion (Costantino et al., 2012). CB₂R agonists also affect the migration of HIV-infected cells: the HIV protein Tat attracts immune cells by activating chemokine receptors, and cannabinoids prevent migration of microglia-like BV2 cells to Tat (Fraga et al., 2011). Direct activation of CB₂Rs also inhibits extracellular matrix adhesion, which monocytes use to cross the blood-brain barrier in response to HIV (Raborn et al., 2014). CB₂R activation could, therefore, slow the spread of HIV infection through microglial populations, and reduce recruitment of peripheral immune cells that may contribute to neuroinflammation.

Despite in vitro evidence that activation of CB₂Rs affords neuroprotection in the presence of HIV or its proteins, few in vivo studies have used drugs selective for CB₂Rs in models of HAND, making it difficult to interpret the contribution of CB₂ versus CB₁Rs. Nevertheless, the few in vivo studies that have specifically investigated CB₂Rs in HAND models suggest that their activation is beneficial. In a humanized mouse model infected with HIV, the CB₂R agonist Gp1a reduced levels of the pro-inflammatory cytokine TNFα and microglial activation, as measured by CD11b expression (Gorantla et al., 2010). In gp120 transgenic mice, the selective CB₂R agonist AM1241 rescued neurogenesis by increasing proliferation and decreasing apoptosis of neural precursor cells (Avraham et al., 2014). While intriguing, these studies did not investigate neuropathology in mature neurons, or assess functional/behavioral outcomes.

Although these treatment strategies have yet to be translated to humans, there is evidence that cannabinoid agonists are well-tolerated. Many HIV patients use cannabis, either recreationally or medically as an appetite stimulant, without severe adverse effects on overall immune function (Pacek et al., 2018). However, there is little evidence that cannabis users are any less susceptible to HAND than non-users. In fact, Δ⁹-THC itself can impair cognitive functioning in both controls and HIV-positive individuals (Thames et al., 2016), although this may occur via its psychoactive effects on CB₁ rather than a CB₂R-mediated effect (Terranova et al., 1995).

Several studies raise concerns about targeting CB₂Rs to treat neurodegenerative disorders such as HAND. While CB₂R activation is largely associated with anti-inflammatory effects, there is evidence suggesting a pro-inflammatory role. Specifically, CB₂R antagonism significantly reduced the release of prostaglandin E2 from primary microglial cells (Saliba et al., 2018), and topical application of 2-AG resulted in dermal inflammation that was blocked by a CB₂R antagonist (Oka et al., 2005; Oka et al., 2006). Furthermore, 2-AG enhances the motility of microglia in part via CB₂R activation, suggesting that these receptors play a role in the accumulation of microglia at sites of neuroimmune insult (Walter et al., 2003). Thus, further identification of the pro-inflammatory responses induced by CB₂R activation will help to define the potentially beneficial role of CB₂R agonists in neuroinflammatory disorders.

Because studies in humans focus on Δ⁹-THC, dronabinol, cannabidiol, or some combination thereof, there is little data on whether a CB₂R-selective compound would fare better in humans. The currently available studies investigating medical marijuana in HIV have been limited by small sample size, short duration of study, and difficulty keeping participants blinded to Δ⁹-THC-containing compounds (Lutge et al., 2013). One CB₂R-selective agonist, APD371, is currently in phase 2 clinical trials for the treatment of abdominal pain in Crohn’s disease (ClinicalTrials.gov Identifier: ); however, this compound has low brain penetrance (Han et al., 2017), which makes it better for treating peripheral diseases rather than HAND. To generate high-quality data on the efficacy of CB₂R agonists in HAND patients, well-controlled trials using selective CB₂R agonists that cross the blood brain barrier are needed.

Despite the aforementioned limitations, selective CB₂R activation remains a promising treatment avenue for two main reasons. First, because CB₂R agonists primarily target immune cells rather than neurons, they have a lower risk of abuse and fewer psychoactive effects than CB₁R agonists or mixed agonists. Second, because CB₂R expression is increased during HIV and SIV infection (Benito et al., 2005; Cosenza-Nashat et al., 2011), CB₂R agonists are most effective when neuroprotection is most needed.

3. Endogenous Ligands, Synthesis, and Metabolism

Several putative endogenous cannabinoids have been discovered (Hanus et al., 1993; Hanus et al., 2001); here, we focus on the two best-characterized ligands anandamide (AEA) and 2-arachidonoylglycerol (2-AG) (Fig. 1). AEA functions relatively slowly and generates both retrograde and non-retrograde signaling (Ohno-Shosaku and Kano, 2014). 2-AG is best known for its role as a retrograde messenger in the CNS, playing an important role in spike timing-dependent plasticity and long-term depression (Kano, 2014). AEA and 2-AG levels can be increased by inhibiting the metabolic enzymes that degrade them. Numerous selective eCB metabolic inhibitors have been developed and extensively characterized in neuropsychiatric and neurodegenerative models. Unlike broad and sustained CB₁ or CB₂R activation produced by agonists, boosting eCB tone by inhibiting metabolic enzymes enhances eCB signaling on-demand and in a site-specific manner, resulting in fewer side effects (Farrell and Soltesz, 2018). Metabolic enzyme inhibitors also overcome the issue of rapid degradation of 2-AG or AEA following direct administration (Savinainen et al., 2001). In this section we will focus on the neuroprotective effects of these modulators, discussing common and unique aspects of their uses and mechanisms in neurological disorders, and their therapeutic potential in HAND.

3.1. Inhibition of AEA metabolism

AEA was first isolated in 1992 from porcine brain in a screen for endogenous ligands for the cannabinoid receptor (Devane et al., 1992). AEA behaves as a partial agonist at both CB₁ and CB₂Rs, and a full agonist at the TRPV1 receptor (Mackie et al., 1993; Sugiura et al., 1999; Sugiura et al., 2000). In the brain, AEA mediates long-term depression (LTD) via the CB₁ and TRPV1 receptors (Grueter et al., 2010; Ohno-Shosaku and Kano, 2014). AEA is an arachidonic acid (AA) derivative, generated from its membrane precursor N-arachidonoyl phosphatidyl ethanolamine (NAPE) through cleavage by a phospholipase D (NAPE-PLD) (Di Marzo, 2008; Liu et al., 2006). Hydrolytic metabolism of AEA occurs via fatty acid amide hydrolase (FAAH) (Cravatt et al., 1996), and oxidative metabolism through various lipoxygenases and cyclooxygenase-2 (COX-2) (Burstein et al., 2000). Pharmacological inhibition of FAAH is a promising approach to enhance AEA function.

Selective FAAH inhibitors have been developed and utilized for nearly 20 years. These drugs fall into several families: α-ketoheterocycles (e.g. OL-135), carbamates (e.g. URB597), alkylsufonylfluorides (e.g. AM3506), and aryl ureas (e.g. PF3845) (Blankman and Cravatt, 2013). Inhibition of FAAH activity blocks the degradation of AEA, and thus holds the potential to enhance AEA-mediated synapse-specific neuroprotection (Farrell and Soltesz, 2018).

FAAH inhibitors exhibit robust analgesic and anxiolytic properties. They show particular promise for alleviating neuropathic pain, including in a model of HIV sensory neuropathic pain in which both CB₁ and CB₂Rs were activated (Chang et al., 2006; Mitchell et al., 2007; Nasirinezhad et al., 2015; Russo et al., 2007). Several novel FAAH inhibitors have recently been developed; these inhibitors are as effective for pain as classical FAAH inhibitors, but with improved metabolic, pharmacokinetic, and brain penetrative properties (Bhuniya et al., 2019). Enhancing AEA levels using FAAH inhibitors also produced anxiolytic effects in a CB₁R-dependent manner in rodents, with few adverse effects (Busquets-Garcia et al., 2011; Gomes et al., 2011; Kinsey et al., 2011a; Kinsey et al., 2011b).

Inhibition of FAAH decreases the severity of seizures (Karanian et al., 2007): the FAAH inhibitor URB597 attenuated limbic seizures in a CB₁R-dependent manner without impairing short- and long-term plasticity (Colangeli et al., 2017). Activation of CB₁Rs is thought to be the primary mechanism of neuroprotection produced by FAAH inhibitors (Schlosburg et al., 2009), although recent studies suggest the TRPV1 receptor may contribute to this protection (Jee Kim et al., 2018; Naziroglu et al., 2018). Repeated seizures are commonly associated with neurodegenerative diseases such as Alzheimer’s disease and traumatic brain injury (Vossel et al., 2017; Webster et al., 2017), and commonly occur in HIV-infected individuals (Ssentongo, 2019). FAAH inhibition is protective in models of Alzheimer’s disease, traumatic brain injury and amyotrophic lateral sclerosis (Bilsland et al., 2006; Tchantchou et al., 2014; Vazquez et al., 2015a). In these disease models, FAAH inhibitors reduce neuroinflammation via cannabinoid receptors, although additional mechanisms may be involved (Vazquez et al., 2015b). Notably, FAAH-mediated degradation of other fatty acid ethanolamide molecules, including palmitoylethanolamide and oleoylethanolamide, alongside AEA may play a role in the potent analgesic and anti-inflammatory effects of FAAH inhibitors (for review see (Alhouayek and Muccioli, 2014)).

Recent clinical evidence indicates several potential safety concerns with FAAH inhibitors. In a phase 1 study, the reversible FAAH inhibitor BIA 10–2474 was orally administered to healthy volunteers to assess safety. Three of four participants developed an acute and rapidly progressive neurologic syndrome, including headache, a cerebellar syndrome, memory impairment, and altered consciousness, and one participant became brain dead (Kerbrat et al., 2016). The underlying mechanism of this unanticipated neurologic disorder remains unknown; however, several off-target effects of BIA 10–2474 were subsequently identified in silico (Molinski et al., 2017) and in vitro (van Esbroeck et al., 2017). The selective FAAH inhibitor PF-04457845 was well-tolerated in humans with no evidence of cannabimimetic adverse events, but lacked efficacy in the primary end point (Huggins et al., 2012). However, this study holds promise that selective FAAH inhibition can be tolerated by humans. Despite the limitations from these and other clinical studies, preclinical data suggest that selective FAAH inhibition may be beneficial in HIV-associated neurodegeneration.

Several recent studies have evaluated inhibition of FAAH in HAND models. Genetic deletion of FAAH rescued neurogenesis in transgenic mice expressing the HIV envelope glycoprotein gp120 (Avraham et al., 2015). The FAAH inhibitor PF3845 also protected against neuronal damage induced by HIV viral proteins via a mechanism involving cannabinoid receptors (Hermes et al., 2018). Thus, FAAH inhibition has therapeutic potential in HAND, although more investigation is needed.

3.2. Inhibition of 2-AG metabolism

2-AG was first isolated as an agonist for cannabinoid receptors in 1995 (Mechoulam et al., 1995). Unlike AEA, 2-AG is present in the brain at high concentrations and is a full agonist for both CB₁ and CB₂Rs (Buczynski and Parsons, 2010; Gonsiorek et al., 2000; Sugiura et al., 1999; Sugiura et al., 2006). It is derived from AA-containing membrane phospholipids, particularly phosphatidylinositol bisphosphate (PIP₂). Phospholipase C (PLC) hydrolyzes the membrane phospholipid into diacylglycerol (DAG), which is cleaved into 2-AG by diacylglycerol lipase (DGL) (Kano et al., 2009; Sugiura et al., 2006). 2-AG production is Ca²⁺-dependent, and it functions as a retrograde signaling molecule that is released postsynaptically after depolarization-induced Ca²⁺ influx or activation of G_αq-coupled metabotropic receptors (Ohno-Shosaku et al., 2012). 2-AG diffuses across the synaptic cleft to stimulate presynaptic CB₁Rs, suppressing the release of neurotransmitters; 2-AG also mediates CB₁R-dependent LTD (Heifets and Castillo, 2009; Ohno-Shosaku and Kano, 2014). Hydrolytic metabolism of 2-AG occurs via monoacylglycerol lipase (MGL) and the serine hydrolases α/β hydrolase domain 6 and 12 (ABHD6 and 12) (Blankman et al., 2007; Di Marzo, 2008). MGL is located in neighboring astrocytes and in the presynaptic terminal, where it hydrolyzes 2-AG into AA and glycerol, thus terminating retrograde signaling. In the brain, 2-AG hydrolysis is the main source of AA for COX-2-mediated production of prostaglandins (Di Marzo et al., 2015; Nomura et al., 2011). Thus, MGL and DGL inhibitors have been explored as an approach to treat neuroinflammation. DGL inhibitors show promise for treating metabolic disorders (Janssen and van der Stelt, 2016), although the psychiatric side effects associated with decreased eCB signaling limit the therapeutic potential of this approach. On the other hand, MGL inhibitors increase 2-AG levels and decrease AA production, thus simultaneously enhancing eCB signaling and decreasing prostaglandin production.

Selective MGL inhibitors were developed slightly later than FAAH inhibitors and include: first-generation MGL inhibitors (e.g. URB602, NAM, and OMDM169), o-aryl carbamates (e.g. JZL184), and o-hexafluorisopropyl carbamates (e.g. KML29) (Blankman and Cravatt, 2013; Granchi et al., 2017). Inhibition of MGL both increases 2-AG levels in the brain, which boosts cannabinoid receptor-mediated neuroprotection and decreases the production of prostaglandins, reducing neuroinflammation (Di Marzo et al., 2015).

Similar to FAAH inhibitors, MGL inhibitors attenuate neuropathic pain and anxiety mainly via CB₁R activation (Crowe et al., 2017; Kamimura et al., 2018; Kinsey et al., 2009; Kinsey et al., 2011a; Kinsey et al., 2011b; Sciolino et al., 2011). MGL inhibitors reduce seizures in several epileptic models, although the underlying mechanism remains contentious (Terrone et al., 2018; von Ruden et al., 2015b). Von Ruden et al. (2015) found that the MGL inhibitor JZL184 attenuated kindling progression in wild-type but not CB₁R knockout mice, suggesting that CB₁Rs suppress epileptogenesis (von Ruden et al., 2015a). However, in a recent study, Terrone et al. (2018) found that JZL184’s therapeutic effect is predominantly anti-inflammatory and not CB₁R-dependent, caused by decreased availability of AA and reduced COX-2-mediated prostaglandin production (Terrone et al., 2018). One possible explanation for this difference is that the latter study used a chronic model of epilepsy, suggesting that the CB₁R-mediated pathway may play a role initially, but that chronic suppression of inflammation maintains the anti-seizure effect long-term. This transient role for CB₁Rs may result from 2-AG overload and desensitization of cannabinoid receptors following excessive inhibition of MGL (Chanda et al., 2010; Schlosburg et al., 2010). Overall, these studies indicate that MGL inhibition is neuroprotective via dual suppression of excitotoxicity and reduced neuroinflammation. A highly potent, orally available CNS-penetrant MGL inhibitor, ABX-1431, is currently in a phase 2 clinical trial for treatment of Tourette Syndrome and a phase 1 trial for neuropathic pain, suggesting therapeutic potential for this class of drugs (Cisar et al., 2018) (ClinicalTrials.gov Identifier: , ).

The effects of MGL inhibition in models of neurocognitive disorders has been extensively studied in recent years, with promising results. MGL inhibitors improved cognition and protected neuronal function in models of Alzheimer’s disease and Down syndrome (Chen et al., 2012; Lysenko et al., 2014). The protection afforded by inhibition of MGL was independent of CB₁R activation, and instead involved neuroinflammatory pathways (Chen et al., 2012; Pihlaja et al., 2015; Piro et al., 2012; Yan et al., 2016) where activation of CB₂Rs on microglia reduced the production of inflammatory cytokines, and the decreased pool of AA reduces prostaglandin production. These studies further suggest that inhibition of MGL might improve cognition in HAND patients.

Recently, we directly tested MGL inhibition in an in vitro model of HAND and demonstrated that the MGL inhibitor JZL184 protects against neuronal damage and excitatory synapse loss induced by HIV viral proteins (Zhang and Thayer, 2018). The protective mechanism involved activation of cannabinoid receptors, as well as decreased AA production. More studies in HAND models in vivo are needed to further evaluate MGL inhibitors in the treatment of HAND.

While MGL functions as the principle enzyme responsible 2-AG hydrolysis (Savinainen et al., 2012), ABHD6 and ABHD12, two other enzymes that hydrolyze 2-AG, play a smaller but still significant role. These enzymes have been evaluated for their role in eCB-mediated neuroprotection (Savinainen et al., 2012). Recent reports indicate that ABHD6 is the major 2-AG hydrolase in cell types not expressing MGL (Marrs et al.), while ABHD12 is more involved in phospholipid rather than 2-AG metabolism (Grabner et al., 2017). Only a limited number of ABHD6 inhibitors have been developed. These inhibitors include carbamate-based inhibitors WWL70, WWL123 and JZP430, and triazole urea-based inhibitors KT195 and KT20. Many of these drugs have significant off-target effects at high concentrations (Patel et al., 2015; Poursharifi et al., 2017); thus, there is a need for development of more potent and selective ABHD6 inhibitors.

The neuroprotective effects of ABHD6 inhibition have been reported in several models of neurological disorders, including neuropathic pain, seizure, traumatic brain injury, and multiple sclerosis. The mechanisms underlying neuroprotection by inhibition of ABHD6 are varied. In a model of neuropathic pain, treatment with the selective ABHD6 inhibitor WWL70 attenuated thermal hyperalgesia and mechanical allodynia by reducing the production of prostaglandin E2 and expression of COX-2, indicating an anti-inflammatory effect (Wen et al., 2018). In a model of multiple sclerosis, ABHD6 inhibition reduced neuroinflammation by activating CB₂Rs (Manterola et al., 2018; Wen et al., 2015). A mouse model of traumatic brain injury reported that WWL70 protects through activation of CB₁Rs (Tchantchou and Zhang, 2013). In contrast to these studies, in a pentylenetetrazole-induced seizure model, ABHD6 inhibition alleviated epileptic symptoms through a cannabinoid receptor-independent pathway that involved GABA_A receptors (Naydenov et al., 2014). Thus, the specific neuroprotective mechanisms of ABHD6 inhibition varies across different disease models and requires further investigation. Inhibition of ABHD6 provides a novel strategy to evaluate in HAND and avoids side effects associated with desensitization of cannabinoid receptors.

3.3. Dual inhibition of AEA and 2-AG metabolism

Dual FAAH and MGL inhibitors attenuate the neuroinflammation and excitotoxicity associated with neurological disorders, such as chronic pain and seizure (Naidoo et al., 2012; Sakin et al., 2015; Sakin et al., 2018; Woodhams et al., 2017). JZL195 is the most commonly used dual inhibitor; other dual inhibitors include organophosphates (e.g. isopropyldodecylfluorophosphonate), and o-hydroxyacetamide carbamates (e.g. SA-57) (Blankman and Cravatt, 2013). The ability of these dual inhibitors to enhance eCB-mediated neuroprotection is unclear. In rodent models of pain, these dual inhibitors usually display higher efficacy than selective MGL or FAAH inhibitors alone, and a wider therapeutic window than cannabinoid receptor agonists (Adamson Barnes et al., 2016; Anderson et al., 2014). However, in an anxiety model, JZL195 did not reduce anxiety, in contrast to selective MGL or FAAH inhibitors (Bedse et al., 2018). Recent reports show that dual inhibitors produce more cannabinoid-related side effects compared to the selective MGL or FAAH inhibitors (Seillier et al., 2014; Wise et al., 2012). Thus, the relative risks and benefits of FAAH-MGL dual inhibitors must be evaluated further in neurological disease models, including HAND.

4. Pathological Impairment of the Endocannabinoid System

One important determinant of clinical outcome is the degree to which the target biological system is functional. Studies in vitro and in vivo have shown that exposure to excitotoxic stimuli alters the eCB system (Chen et al., 2003; Feng et al., 2016; Li et al., 2012) with increased CB₁R-mediated inhibition of GABA release changing the balance of network excitability. Currently, it is not known if eCB signaling is altered in the presence of HIV. In the context of another neurodegenerative disorder, Westlake et al. (1994) found reduced CB₁R agonist binding in the hippocampus and caudate of Alzheimer’s patients using in vitro receptor autoradiography. Using in situ hybridization histochemistry, they also found regionally discrete losses of mRNA expression in Alzheimer’s relative to normal brains (Westlake et al., 1994). Similarly, Ramirez et al. (2005) determined that cannabinoid receptors couple less efficiently to their G-proteins in Alzheimer’s brains (Ramirez et al., 2005). These and other studies suggest that impaired CB₁R-mediated signaling might diminish the neuroprotective capacity of the eCB system, reducing the efficacy of exogenous CB₁R agonists.

Several reports indicate changes in CB₁ and CB₂Rs during SIV and HIV infection. Benito et al. (2005) identified increases in CB₂R immunoreactivity in cortical tissue from rhesus macaques with SIV-induced encephalitis (Benito 2005). In this same study, these SIV-infected macaques also showed increased FAAH immunoreactivity in brain cortical samples. These results agree with previous observations in Alzheimer’s disease brains (Benito et al., 2003), indicating a common pattern of response despite different primary inflammatory insults. Indeed, Consenza-Nashat et al. (2011) found CB₁ and CB₂R upregulation in macrophages and microglia in tissue from patients with HIV encephalitis (HIVE). The functional consequences of these changes are unknown; thus, it is important to determine how eCB signaling changes during exposure to HIV to evaluate the feasibility of targeting components of the eCB system and, if necessary, devise strategies to protect or enhance signaling.

5. Conclusions and Perspectives

The unique pathophysiology of HAND may be particularly amenable to modulation by the eCB system. The excitotoxic component of HAND is attenuated by activating presynaptic CB₁Rs, either directly by agonists or indirectly through inhibition of eCB metabolism. The chronic neuroinflammation associated with HAND can be attenuated by activating CB₂Rs on immune cells or inhibiting MGL-mediated AA production. Furthermore, expression of cannabinoid receptors is increased in HAND, potentially enhancing the sensitivity of the eCB system in the presence of HIV.

There are drawbacks to targeting certain elements of the eCB system. Of particular concern are the psychoactive effects resulting from CB₁R activation. Motor impairment and reduced short term memory limit the utility of CB₁R agonists. However, it may be possible to avoid the desensitization of cannabinoid receptors that accompanies chronic activation by enhancing endogenously produced agonists with MGL and FAAH inhibitors. Because HAND patients will likely not be treated prophylactically, it will be essential to understand the time course of the disease, particularly changes in neuroinflammation, to understand the relative importance of excitotoxic and neuroinflammatory targets.

In summary, we have discussed several promising targets in the eCB system for suppressing the neuronal damage that underlies HAND (Fig. 1). CB₂R agonists suppress neuroinflammation without the psychoactive side effects of non-selective drugs that also activate CB₁Rs. MGL inhibitors are promising because they potentiate endogenously produced 2-AG to inhibit glutamate release via CB₁Rs and suppress immune function via CB₂Rs, while also decreasing the pool of AA needed for the synthesis of prostaglandins in the CNS. Both CB₂R-selective agonists and MGL inhibitors are well-tolerated and are currently in clinical trials for indications other than HAND. The work reviewed here suggests that these agents may slow the progression of HAND in individuals living with HIV.

Highlights.

• Nearly half of all HIV-infected individuals experience cognitive and motor deficits.

• HIV-induced neuronal injury results from excitotoxic and inflammatory mechanisms.

• The endocannabinoid (eCB) system provides on-demand protection against excitotoxicity and neuroinflammation.

• We discuss the potential of drugs that modulate eCB signaling to treat HIV-associated neurocognitive disorder.

Abbreviations

Δ⁹-THC

Δ⁹-tetrahydrocannabinol

2-AG

2-arachidonoylglycerol

AA

arachidonic acid

AEA

arachidonoyl ethanolamine (anandamide)

AIDS

acquired immunodeficiency syndrome

CB₁R

cannabinoid type 1 receptor

CB₂R

cannabinoid type 2 receptor

DAG

diacylglycerol

DGL

diacylglycerol lipase

FAAH

fatty acid amide hydrolase

GPCR

G protein-coupled receptor

HIV

human immunodeficiency virus

IP₃

inositol triphosphate

MGL

monoacylglycerol lipase

NAPE-PLD

N-arachidonoyl phosphatidylethanolamine phospholipase D

NAT

N-acyltransferase

NAPE

N-arachidonoyl phosphatidylethanolamine

PE

phosphatidylethanolamine

PIP₂

phosphatidylinositol 4,5-bisphosphate

PLC

phospholipase C

SIV

simian immunodeficiency virus

VGCC

voltage-gated Ca²⁺ channel