Chemical Tools for Studying Lipid-Binding Class A G Protein–Coupled Receptors

1. Radioligands for In Vitro Experiments.

Rather than an exhaustive commentary of all radioligands and uses, this review describes the commonly used in vitro radioligands based on synthetic scaffolds (Fig. 1; Table 1) and describes selected studies that showcase the utility of these radioligands. It should be noted that many radiolabeled lipids are commercially available and have been used to study receptor and endogenous ligand trafficking, for example [³³P]S1P (Jo et al., 2012), [³H]LPA (Thomson et al., 1994), [³H]AEA (Hillard et al., 1997), [³H]PGE₂ (Nemoto et al., 1997), [³H]PGD₂ (Matsuo and Cynader, 1993), [³H]LTB₄ (Toda, 1999), [³H]5-oxo-ETE (O'Flaherty et al., 1998), and [³H]PAF (Nakamura et al., 1991). Although these labeled endogenous ligands can be useful tools for interrogating receptor pharmacology (Hecht et al., 1996; Parrill et al., 2000; Van Brocklyn et al., 2000), they often suffer from rapid enzymatic degradation even in in vitro experiments. In this study, we review only radiolabeled synthetic ligands.

In 1964, THC was identified as a CB₁ receptor ligand, and soon after it was ¹⁴C-labeled to provide the first CB receptor radioligand [¹⁴C]THC (Miras, 1965). Since then a number of THC-based CB receptor radioligands have been developed (Agurell et al., 1969; Pitt et al., 1980; Nye et al., 1985); however, nowadays these are not commonly used due to high nonspecific membrane binding.

Autoradiographic studies of receptor distribution can be carried out in vitro; for example, an early report using the CB receptor radioligand [³H]CP-55,940 (Fig. 1; Table 1) found the highest CB₁ receptor density in substantia nigra, basal ganglia, globus pallidus, cerebellum, and hippocampus regions that are known to control cognitive and motor functions (Herkenham et al., 1990; Glass et al., 1997). An interesting recent example of a competition-binding assay using [³H]CP-55,940 is the evaluation of approximately 50 compounds from the herbal mixture known as “Spice,” previously shown or assumed to be a CB₁ agonist (Hess et al., 2016). Many new structural analogs of cannabinoids have emerged in recent years designed to circumvent legally defined chemical classes, with little accompanying pharmacological evaluation. Hess et al. (2016) carried out radioligand-binding assays using [³H]CP-55,940, which revealed most compounds had low to subnanomolar affinity for CB₁ and CB₂ receptors, and further functional assays showed these to be agonists.

CB receptor affinity determined by a radioligand competition-binding assay is often the first parameter by which ligands are ranked despite the sometimes little structural similarity between radioligand and test ligand. A recent report has highlighted some important things to note when using a universal radioligand for investigating binding of ligands to CB₂ receptor (Smoum et al., 2015). Smoum et al. (2015) reported that HU-433, the enantiomer of the well-studied HU-308, had 25-fold reduced binding affinity to CB₂ receptor (as measured using a competition assay with [³H]CP-55,940) but had much higher potency in osteoblast proliferation and anti-inflammatory experiments compared with HU-308. It was proposed that HU-433 and HU-308 may bind to CB₂ receptor in different orientations, thus turning on different downstream signaling pathways and competing with [³H]CP-55,940 nonequally. This study serves as a reminder that measured compound affinity for a receptor is only ever in relation to the reference, and sometimes the choice of reference or radioligand tool can influence not just the raw number but also the rank order of ligands.

Radioligands are crucial tools for investigating the existence of multiple GPCR states. For example, it was estimated that CB receptors are present in a 70% inactive:30% active conformation in rat cerebellar membrane using the agonist [³H]CP-55,940 that bound only to active rat (r) CB1 receptor and the inverse agonist [³H]SR141716A (Fig. 1; Table 1) that bound to rCB₁ receptors in both active and inactive states (Kearn et al., 1999). Radioligands are also important tools for identifying key receptor amino acids, for example, direct comparison of radioligand binding to wild-type versus mutant receptor. In one study, modified human (h) CB₁ receptors were constructed with residues D2.63 and K373 mutated to Ala or the reciprocal Asp/Lys mutation to investigate the importance of an ionic interaction (Marcu et al., 2013). [The residue numbering system used is based on the Ballesteros–Weinstein numbering scheme, whereby the single-letter code of the amino acid is followed by the transmembrane helix (TMH) number, and then the amino acid’s position relative to the TMH’s most conserved residue (assigned as 50)]. The K_d, B_max, and K_i of [³H]SR141716A were measured for wild-type receptor and for each receptor mutant, and, interestingly, no significant difference in binding was observed. However, a GTPγS-binding assay indicated signaling was impaired in each Ala receptor mutant, but not in the reciprocal Asp/Lys mutant, with the authors suggesting the ionic interaction between D2.63 and K373 influences the EC-3 loop and is important for CB receptor signaling.

The physiologic relevance of GPCR oligomerization is not yet fully understood, and radioligands have been used as tools to shed light on this; for example, a selective adenosine_2A receptor agonist CGS21680 decreased the B_max of the CB₁ receptor radioligand [³H]SR141716A in synaptosomal membrane of rat brain from 3.23 ± 0.17 pmol.mg⁻¹ to 2.68 ± 0.18 pmol.mg⁻¹ (Ferreira et al., 2015). This was attributed to inhibition of CB₁ receptor signaling by adenosine_2A receptor, with the authors suggesting that dual targeting of these receptors may prove therapeutically beneficial. Despite radioligands having utility for investigating GPCR oligomerization, fluorescent tools are often superior for this line of investigation, as bioluminescence resonance energy transfer (BRET) and fluorescence resonance energy transfer experiments can provide precise spatial information (Applications of Small-Molecule Fluorescent Tools). Allosteric binding sites of CB, FFA, prostanoid, and S1P receptors have been interrogated using radioligand tools. In one example, a CB₁ receptor radioligand-binding assay carried out in the presence of compounds such as ORG27569 (Fig. 6) increased the binding affinity of the agonist [³H]CP-55,940, whereas a decrease in the binding affinity for the inverse agonist [³H]SR141716A was observed (Price et al., 2005).

Distinct FFA1 receptor binding sites have been identified using the partial allosteric agonist [³H]AMG837 (position of the ³H radioisotope was not reported) and the full allosteric agonist [³H]AM1638 (Table 1) in experiments with and without endogenous FFA docosahexaenoic acid and other unlabeled FFA1 receptor agonists (Lin et al., 2012). Three allosterically linked binding sites were proposed, with the authors noting these ligand–receptor interactions were far more complex than previously recognized. These findings are important in the context of the role of FFA1 receptor in glucose-stimulated insulin secretion and may provide a novel multiligand treatment therapy for type 2 diabetes, in which a reduced dose of drug can be used due to ligand cooperativity. Another research group has reported the synthesis and use of [³H]TAK-875 (Fig. 1; Table 1) as a FFA1 receptor radioligand agonist (Bertrand et al., 2016a), based on the TAK-875 drug that failed Phase II clinical trials. After demonstrating that [³H]TAK-875 bound to GPR40–human embryonic kidney (HEK) cells in higher levels than in wild-type HEK cells, the research team developed a [¹⁸F]TAK-875 radioligand (Radioligands for In Vivo Experiments).

The first reported radioligand-binding assay for FFA2 receptor used [³H]GLPG0974 (Table 1; position of the ³H radioisotope not reported) as the competing ligand (Sergeev et al., 2016). This radiolabeled antagonist was characterized using saturation equilibrium-binding assays across several different FFA2 receptor mutants and then analyzed in kinetic binding assays, in particular to determine on and off rates of unlabeled agonists and antagonists. Data from these experiments led to the suggestion that an agonist requires interaction with both arginine residues (Arg-180^5.39 and Arg-255^7.35), whereas, for an antagonist, interaction with only one of these two arginines is required. This provided valuable information to aid rational structure-based design of novel tools and drugs for this therapeutically relevant receptor. Radioligands based on synthetic, selective scaffolds for FFA3 and FFA4 receptors have not yet been reported; however, researchers have used labeled fatty acids such as [³³P]S1P to probe the binding site(s) of unlabeled synthetic agonists (Jo et al., 2012).

To eventually develop iodinated derivatives for in vivo imaging of S1P receptor (Radioligands for in vivo experiments), Briard et al. (2015) made two radioligands—[¹⁴C]BAF312 and a cold iodinated derivative of this [¹⁴C]18 (Fig. 1; Table 1) (Briard et al., 2015). These two ligands were used in autoradiography studies to compare biodistribution of the MS drug BAF312 (Fig. 1) to the iodinated derivative [¹²⁷I]MS565 with the goal of making [¹²³I]MS565. Whole-body autoradiography studies in rats revealed similar biodistribution profiles of [¹⁴C]BAF312 and [¹⁴C]18.

Radioligands based on synthetic LPA receptor ligands are not yet reported despite the availability of several high-affinity synthetic ligands for these receptors (Ohta et al., 2003; Qian et al., 2012). It is expected that development of radioligands based on synthetic ligands for studying LPA receptors will see rapid growth. This development will be fueled by the need to prepare novel ligands targeting these receptors for various disease conditions.

Selective synthetic radioligands are available for prostanoid family DP₁, DP₂, FP, and TP receptors. The selective DP₁ receptor antagonist [³H]BWA868C (Table 1) has been used for autoradiographic studies in human eye sections (Sharif et al., 2000). This study revealed high levels of DP₁ receptor expression in ciliary epithelium/processes, ciliary muscles, the iris, and the retinal choroid, which was similar to that observed previously using the endogenous ligand [³H]PGD₂ (Matsuo and Cynader, 1993). Selective radiolabeled DP₂ receptor antagonists have been reported: [³H]TRQ11238 (Ulven et al., 2007), [³H₃]QAW039 (Luu et al., 2015), and [³H]OC-459 (Sykes et al., 2016) (Fig. 1; Table 1). Consideration of ligand–receptor kinetics is crucial in a drug discovery program, and often drug dissociation time from the receptor is clinically relevant. Sykes et al. (2016) therefore used [³H₃]QAW039 and [³H]OC-459 (tritiated versions of the Phase III asthma treatments fevipiprant and timapiprant, respectively) to directly measure drug kinetics and function in parallel to unravel the drug mechanism of action, concluding that QAW039 should competitively inhibit disease-relevant DP₂ receptor-mediated responses in human cells (Sykes et al., 2016).

Sharif et al. (1999) used the selective FP receptor agonist [³H]9β‐(+)‐fluprostenol (Fig. 1; Table 1), a carboxylic acid derivative of the glaucoma drug travoprost, to study prostanoid receptor distribution in the human eye (Sharif et al., 1999), which revealed FP receptor distribution consistent with that obtained using tritium-labeled endogenous ligand [³H]PGF_2α (Matsuo and Cynader, 1992; Davis and Sharif, 1999).

[³H]Iloprost (Fig. 1; Table 1), the tritiated analog of the marketed agonist drug iloprost, has been used to investigate IP receptor expression and function, although it also possesses modest potency for EP₃ and EP₁ receptors (Abramovitz et al., 2000). Decreased saturation binding of [³H]iloprost in platelets of humans with type 2 diabetes compared with healthy humans has been demonstrated (Knebel et al., 2015). This finding along with other functional measurements led the authors to suggest that this lower IP receptor expression in platelets of patients with type 2 diabetes might lead to increased platelet aggregation, which may be the reason for the increased risk of thrombosis observed in such patients.

Among prostanoid receptors, the greatest number of radioligands has been developed to study TP receptor (Fig. 1; Table 1), probably due to its importance in thrombosis and related cardiovascular disorders. Selective antagonist [³H]SQ-29548 (Hanasaki et al., 1988; Hedberg et al., 1988) (Fig. 1; Table 1) is frequently used in competition-binding assays. For example, in one study, the antibody C-EL2Ab, which binds to the C-terminal of second extracellular loop of TP receptor, competitively inhibited the binding of [³H]SQ-29548 to TP receptor and also inhibited TP receptor–mediated platelet aggregation, which showed the importance of this portion of the receptor structure in platelet activation and the potential of C-EL2Ab as an alternative antiplatelet agent (Murad et al., 2012).

Compared with the prostanoid receptor family, there are far fewer synthetic scaffold-based radioligands reported for the leukotriene receptor family, and therefore radiolabeled endogenous ligands are frequently used. There is one report of the synthesis and use of the antagonist [³H]CGS23131 (Fig. 1; Table 1) to characterize the single LTB4 receptor subtype recognized at that time on human polymorphonuclear neutrophils (Jackson et al., 1992) and several early reports on the use of [³H]ICI198615 (Fig. 1; Table 1) (Aharony et al., 1988; O'Sullivan and Mong, 1989) to study the single recognized CysLT receptor. Pranlukast, a selective antagonist for CysLT₁ receptor that is used clinically to treat asthma, has been tritium labeled and used for autoradiographic studies in human nasal inferior turbinates (Shirasaki et al., 2006). These studies revealed high distribution of CysLT₁ receptor in vascular endothelium and the interstitial cells; however, the K_d of [³H]pranlukast was not reported.

There are many examples of synthetic high-affinity antagonists for PAF receptor, several of which have been radiolabeled and used as tools, for example [³H]-dihydrokadsurenone (Hwang et al., 1986), [³H]52770-RP (Robaut et al., 1987; Marquis et al., 1988), [³H]WEB2086 (also known as [³H]apafant) (Ukena et al., 1988), [³H]L-659989 (Hwang et al., 1989) (Fig. 1; Table 1), and [³H]SR27417 (Herbert, 1992) (Table 1). Determination of the degree of competition that a synthetic radioligand has with the endogenous receptor agonist is important; for example, it was shown that [³H]WEB2086, but not [³H]52770-RP, interacted with the same binding site as [³H]PAF in human platelets (Ukena et al., 1988). Now commercially available, [³H]WEB2086 has been used several times as a tool to study PAF receptor; for example, radioligand-binding assays showed that human B lymphoid cell line LA350 expressed high levels of PAF receptor (Zhuang et al., 2000). To date, synthetic radiolabeled ligands for GPBA receptor have not been reported; however, it is likely this will be an area of future interest as drugs targeting this receptor (Xu, 2016) are developed.

2. Radioligands for In Vivo Experiments.

In vivo radioligand experiments require consideration of blood brain barrier (BBB) permeability (if required) and ligand metabolism to radiometabolites that can complicate analysis. As discussed in Introduction, selectivity of a chemical tool in vivo requires careful consideration of receptor subtype selectivity (which is often first measured in vitro), versus relative in vivo expression levels of the subtypes/other receptors. Much effort has been directed toward the development of in vivo radioligands to study CB₁ receptor in brain to investigate pain and addiction pathways. Despite PET radioisotopes having a shorter t_1/2 (¹¹C t_1/2 = 20 minutes, ¹⁸F _t1/2 = 110 minutes), Miller et al. (2008) compared with those used in SPECT (¹²³I terminal t_1/2 = 13 hours; γ ray = 159 keV) (Kung et al., 2003). PET has been used extensively to study CB receptors due to higher resolution and sensitivity. A major challenge is that the short radioligand t_1/2 requires on demand synthesis with fast purification. PET and SPECT imaging time, which is also related to radioligand t_1/2, is usually insufficient to allow radioligand–receptor binding to reach equilibrium; therefore, appropriate kinetic models should be used to correct this. In vivo imaging of the ECB system using PET (Horti et al., 2014), brain imaging of CB receptors using PET and SPECT (Casteels et al., 2013), and synthesis of such radioligands (Ahamed et al., 2013) have recently been extensively reviewed. Of note and described in these reviews are radioligands [¹¹C]OMAR, [¹¹C]MePPEP, [¹⁸F]FMPEP-d2, and [¹⁸F]MK-9470, which have been used in preclinical imaging of CB receptors in humans. Since the publication of these reviews, additional novel radioligands have been reported and are discussed in this work along with more recent applications of commonly used in vivo radioligands.

[¹¹C]OMAR (Table 2) is a structural analog of SR141716A (Fig. 1), in which a methyl and chlorine of SR141716A have been replaced with cyano and ¹¹C-labeled methoxy to increase polarity (Fan et al., 2006). [¹¹C]OMAR has been used in human PET studies to characterize hCB₁ receptor expression in cannabis dependence (D'Souza et al., 2016), post-traumatic stress disorder (Neumeister et al., 2013), and threat perception in trauma (Pietrzak et al., 2014). [¹⁸F]MK-9470 (Table 2) is a high-affinity inverse agonist for CB₁ receptor (Burns et al., 2007), which to date is the most widely used radioligand for in vivo CB₁ receptor studies. Since the area was last reviewed (Horti et al., 2014), [¹⁸F]MK-9470 has been used in PET experiments to analyze hCB₁ receptor expression in schizophrenia (Ceccarini et al., 2013), prostate carcinoma (Emonds et al., 2013), Alzheimer's disease (Ahmad et al., 2014), alcohol dependence (Ceccarini et al., 2014), cannabis dependence (Ceccarini et al., 2015), functional dyspepsia (Ly et al., 2015), food intake disorders (Ceccarini et al., 2016b), and Huntington’s disease (Ceccarini et al., 2016a).

The CB₂ receptor in vivo radioligand toolbox is underdeveloped compared with that of CB₁ receptor, perhaps due to the timeline of receptor subtype characterization and the intense interest in CNS-expressed CB₁ receptor. The 2-oxoquinoline–based [¹¹C]NE40 (Table 2) (Evens et al., 2009) was the first radioligand used for in vivo PET imaging of hCB₂ receptor (Ahmad et al., 2013) and will be an important tool for studying CB₂ receptor in pathologic conditions. Ahmad et al. (2013) showed rapid brain uptake and washout of [¹¹C]NE40, along with major uptake in lymphoid tissue in agreement with known expression of hCB₂ receptor.

Hortala et al. (2014) developed a triazine-based radioligand [¹⁸F]d₂-3 (Fig. 2; Table 2), which was used for PET imaging of CB₂ receptor in rhesus monkey and baboon models of neuroinflammation (Hortala et al., 2014). [¹⁸F]5 (Fig. 2; Table 2) is another triazine-based radioligand, which was developed to study in vivo distribution of a series of triazine-based CB₂ receptor agonists (Yrjölä et al., 2015). Other radioligands with an oxoquinoline scaffold have been reported and used for imaging CB₂ receptor in rats and mice, for example, [¹¹C]KD2 (Mu et al., 2013), [¹¹C]RSR-056 (Slavik et al., 2015a), [¹¹C]RS-016 (Slavik et al., 2015b), [¹⁸F]RS-126 (Slavik et al., 2016), and [¹¹C]KP23 (Mu et al., 2014) (Fig. 2; Table 2). These oxoquinoline-based radioligands all exhibited similar patterns of biodistribution with high specific binding to spleen (concluded from excess cold ligand–blocking experiments) and low brain uptake, along with high nonspecific binding to liver and small intestine. It is unclear whether low brain uptake is solely a reflection of proportionally lower CB₂ receptor expression in brain compared with spleen or also a function of poor BBB passage. In the case of [¹¹C]RSR-056 and [¹¹C]RS-016, an increase in brain radioligand uptake in a neuroinflammatory mice model was observed as compared with healthy mice. However, it remains ambiguous whether this increase is a result of the increased expression of CB₂ receptor or a result of increased brain permeability due to inflammation-induced disruption of the BBB.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

CB receptor radioligand tools for in vivo studies reported since (Casteels et al., 2013; Horti et al., 2014). Note: structures are not shown for radioligands with unknown radioisotope position.

Both cis and trans isomers of the naphthyridin-based radioligand [¹⁸F]CB91 were evaluated (Fig. 2; Table 2) with the cis/trans mixture then used for imaging CB₂ receptor in mice and showed distribution in spleen and gut, consistent with known CB₂ receptor distribution, but also in kidneys, pancreas, and brown adipose tissue. High liver accumulation of radioligand indicated rapid clearance by the hepatobiliary route (Saccomanni et al., 2015). [¹¹C]AZD1940 (Fig. 2; Table 2) is a radiolabeled analog of a peripherally restricted CB receptor agonist AZD1940, which was a drug investigated for treatment of neuropathic pain. [¹¹C]AZD1940 was used for PET studies in monkeys to study the distribution of AZD1940 and revealed low CNS exposure; however, further experiments are required to determine the contribution of unbound versus CB receptor–bound ligand in brain given the nanomolar CB₁ receptor affinity of [¹¹C]AZD1940 (Schou et al., 2013). Synthesis of PET radioligands based on aminoalkyl indole (Gao et al., 2014a) and benzenesulfonamide (Gao et al., 2014b) scaffolds has been reported; however, biologic data for these ligands are not yet reported. Most reported CB₂ receptor radioligands exhibit high lipophilicity and metabolic susceptibility or lack CB₂ receptor specificity, and much work still needs to be done to enable intricate in vivo imaging of CB₂ receptor in humans.

As already discussed, there are very few reported FFA receptor radioligands. In the same body of work as [³H]TAK-875 (Radioligands for In Vitro Experiments), Bertrand et al. (2016a) also reported the synthesis of [¹⁸F]TAK-875 (Fig. 3; Table 2). The cold analog [¹⁹F]TAK-875 was also synthesized and showed similar FFA1 receptor agonist activity to TAK-875 (Fig. 1), and in vitro and in vivo studies are underway.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

FFA, S1P, and prostanoid receptor radioligand tools for in vivo studies. Note: structures are not shown for radioligands with unknown radioisotope position. The structures of [^99mTc]RP517 and [¹¹¹In]-DPC11870-11 are not shown because of size.

There are a greater number of in vivo radioligands for S1P compared with FFA and LPA receptors, a trend perhaps reflective of the number of marketed and under development drugs for S1P receptor. The synthesis and evaluation of FTY720 analogs with a cold ¹²⁷I appended to varying positions as a lead-in to making [¹²³I] or [¹²⁴I] radioligands for SPECT or PET S1P receptor imaging have been reported (Briard et al., 2011). The 2-iodo derivative, BZM055, was identified as a lead candidate for radiolabeling, and the authors comment that imaging studies are ongoing. Briard et al. (2015) also took a similar approach and synthesized an iodinated analog of the S1P receptor ligand BAF312 (Fig. 1). Distribution of cold analog [¹²⁷I]MS565 was analyzed using the equivalent [¹⁴C]18 tool (Radioligands for In Vitro Experiments) and is suitable for development of hot [¹²³I]MS565 as a SPECT tool (Briard et al., 2015).

Cold fluorinated derivatives of the S1P₁ receptor antagonist W146 have been synthesized, with the most promising in vitro derivative radiolabeled to [¹⁸F]24 (Fig. 3; binding affinity not reported) (Prasad et al., 2014). Although stable in serum, in vivo PET imaging of S1P₁ receptor in mice using [¹⁸F]24 showed accumulation of radioactivity in bone, most likely due to metabolic defluorination, thus limiting its use. The same laboratory subsequently reported [¹⁸F]17, a derivative of FTY720, for imaging S1P receptors (Fig. 3; Table 2) (Shaikh et al., 2015). [¹⁸F]17 induced peripheral blood lymphopenia (a measure of S1P₁ receptor downstream regulation) at a comparable level to FTY720 and did not undergo defluorination during mice PET studies; however, rapid clearance of [¹⁸F]17 was observed. Thus, further optimization of these PET radioligands is required.

[¹¹C]TZ3321 (Fig. 3; Table 2), based on a selective S1P₁ receptor oxadiazole scaffold, has been reported and used in MicroPET imaging studies in a femoral artery wire–injury mouse model (Jin et al., 2017). These studies revealed high expression of S1P₁ receptor, in line with the proposal that higher levels of this receptor are observed in vascular smooth muscle cells following intimal lesions usually caused by in-stent restenosis. The same research group then reported synthesis of the PET radioligand [¹⁸F]28c (Fig. 3; Table 2), also based on oxadiazole scaffold (Rosenberg et al., 2016). S1P₁ receptor expression is thought to increase in inflammation, and in vivo PET imaging indeed showed increased binding of [¹⁸F]28c in the liver of lipopolysaccharide-treated mice compared with control mice. This radioligand could be an important tool for monitoring S1P₁ receptor expression as a measure of inflammation.

Radioligand tools have been developed for imaging S1P₂ receptor, for example, [¹¹C]5a (Fig. 3; Table 2) (Yue et al., 2015), which is a derivative of the S1P₂ receptor antagonist JTE-013. PET studies in mice showed high uptake of [¹¹C]5a in S1P₂ receptor–rich regions such as heart, lung, kidney, and liver. Low brain levels were observed, with the authors suggesting that this could be due to radioligand efflux by P-glycoprotein (P-gp), as an increase in brain penetrability was observed on treatment with P-gp inhibitor cyclosporine A. Further improvements in brain uptake of this radioligand are required before utility in PET imaging of MS and neuroinflammation conditions is achieved. S1P₃ receptor radioligands have also been developed for in vivo imaging. [¹⁸F]1, a mixture of two compounds (Fig. 3), was synthesized based on a known [¹⁹F]-containing indole-based ligand for S1P₃ receptor (Rokka et al., 2013); however, to date pharmacological data for this radioligand have not been reported.

Four [¹⁴C]-labeled DP receptor antagonists, including 1-[¹⁴C] and 2-[¹⁴C] (Fig. 3; Table 2) (Berthelette and Wang, 2007), along with three [³H]-labeled analogs (Scheigetz et al., 2004), have been synthesized and used to study the extent of covalent protein labeling in vitro (Sturino et al., 2007). These [¹⁴C]-labeled DP receptor antagonists may prove useful tools for in vivo DP receptor studies in the future, although DP receptor selectivity in particular over TP receptor will require further optimization. [¹⁴C]‐Fluprostenol has been used to analyze the disposition and metabolism of an intramuscular injection of fluprostenol in horses (Chapman et al., 1980), although DP receptors were not directly investigated.

An optimized [^99mTc]RP517-containing formulation has been developed as a [^99mTc]-labeled analog of the hydrazinonicotinamide-conjugated LTB4₁ receptor antagonist SG380 (Table 1; unchelated ligand) for imaging infection and inflammation (Liu et al., 2002). A study by Riou et al. (2002) examined [^99mTc]RP517 uptake in an ischemia-reperfusion–induced myocardial inflammation model, and found that postreperfusion [^99mTc]RP517 uptake correlated with myeloperoxidase (a specific neutrophil enzyme) levels. Ex vivo imaging of heart slices postreperfusion also showed [^99mTc]RP517 localized within the area of inflammation. To investigate whether these results might be LTB4₁ receptor specific, Riou et al. (2002) also performed in vitro studies using [F]-RP517, an undisclosed CY3 fluorescently tagged analog of RP517, which showed a 44% displacement with LTB₄ and complete displacement with nonfluorescent RP517 on isolated neutrophils. [^99mTc]RP517 has also been used as a tool to study Escherichia coli infection in rabbits; however, although accumulation of [^99mTc]RP517 occurred in the abscess, there was also significant accumulation in the intestines, leading the authors to propose development of a more hydrophilic tool (Brouwers et al., 2000). This research group then reported ¹¹¹In labeling of the more hydrophilic LTB4₁ receptor antagonist DPC11870-11 (Van Eerd et al., 2003). [¹¹¹In]-DPC11870-11 showed specific receptor interactions, localized to infection foci rapidly after injection and showed minimal accumulation in the intestines.

[¹⁴C]WEB2086 (also called [¹⁴C]apafant) (Table 2) was first reported in a patent in 1989 (Birke and Stiasni, 1989) and has since been used to show apafant is a substrate for P-gp but has not been used to directly interrogate PAF receptor (Leusch et al., 2002; Fuchs et al., 2014). Again, in a pharmacokinetic rather than direct PAF receptor study, [¹⁴C]E6123 (Table 2) has been reported and used to probe metabolic enzymes of PAF antagonist E6123 (Kusano et al., 1993).

PET radioligands are valuable tools for studying the pharmacokinetics of investigational drugs and in vivo imaging of GPCRs. In particular, many in vivo radioligands have been reported for CB receptor, and it is anticipated in the near future more radioligands for the other class A lipid-binding GPCRs will be developed, thus facilitating an understanding of the role these receptors have in disease conditions.

1. Fluorescent Tools for In Vitro Experiments.

Early reports of fluorescent small molecules for lipid-binding GPCRs were of labeled endogenous receptor agonists, for example, a benzoxadiazole-based fluorophore linked to anandamide (Koga et al., 1995) or to S1P (Hakogi et al., 2003; Yamamoto et al., 2008). Bile acid derivatives have been linked to nitrobenzoxadiazole (NBD) fluorophores and used to study cellular uptake (Májer et al., 2012), and fluorescent LTB₄ linked to two AlexaFluor fluorophores has been used specifically to study LTB4₁ receptor (Sabirsh et al., 2005). An early report of a flow cytometry assay for FFA1 receptor used the commercially available fluorescent ligand C1-BODIPY-C12 (Hara et al., 2009). Many endogenous ligands for CB, FFA, and S1P receptor are now commercially available attached to a range of fluorophores.

In 2008, a biotin–AEA tool was developed to study the biodistribution of AEA, which was visualized in a two-step process using an anti-biotin monoclonal antibody (mAb), followed by a green fluorescent anti-mouse secondary antibody (Fezza et al., 2008). Martin-Couce et al. (2011) described modification of the ECBs AEA, 2-AG, and 2-arachidonyl glyceryl ether (2-AGE) to contain a biotin or alkyne tag for subsequent direct, selective in situ fluorescent labeling with a streptavidin-fluorophore or a click reaction, respectively. The 2-AGE ECB tagged with an alkyne was the most potent tool, but it lacked receptor subtype selectivity (K_i = 84.7 ± 0.8 nM at hCB₁; 84.9 ± 0.6 nM at hCB₂), whereas 2-AGE-biotin-3b (Fig. 4; Table 3) showed some selectivity (K_i = 221 ± 8 nM at hCB₁, 450 ± 11 nM at hCB₂). Imaging of hCB₁ receptor in mouse hippocampal cell lines (HT-22) transfected with CB₁ receptor has been carried out by reaction of 2-AGE-biotin-3b with streptavidin-Alexa488 fluorophore in situ. Some background fluorescence was observed in control experiments using either nontransfected cells or with a high concentration of the CB₁ and CB₂ receptor agonist HU210, indicating some nonspecific binding. The ligand 2-AGE and other ECBs are lipids, so using these as a basis for chemical tools means large amounts of nonspecific membrane binding are probable along with modest, if any, CB receptor subtype selectivity, driving the need for fluorescent, synthetic ligands with improved selectivity and physicochemical properties.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

CB receptor fluorescent tools.

The first reported fluorescent tools based on a nonendogenous CB pharmacophore were dansyl derivatives of THC (Forrest et al., 1971), and these were used for analytical tracing of cannabinoid ligands (Just et al., 1972). However, as this was prior to CB receptor characterization, these fluorescent compounds were not evaluated explicitly as CB receptor ligands. About 30 years later, a fluorescent naphthoyl indole CB₂ receptor agonist designed around JWH-015 was developed using a CB₂ receptor homology model and ligand docking that indicated a naphthoyl-linker position may be tolerated and the NBD fluorophore could act as a second pharmacophore with the receptor (Yates et al., 2005). Unfortunately, this fluorescent compound had significant loss of affinity (10 µM gave 25% displacement of [³H]CP-55940 from CB₂ receptor) compared with parent indole agonist JWH-015, cytosolic accumulation, and nonspecific binding. In another report, replacement of part of an orthosteric pharmacophore with a fluorescent moiety has been reported to generate a fluorescent CB₂ receptor tool (Petrov et al., 2011). The morpholine of isatin acylhydrazone, believed to interact with the lipophilic cavity of CB₂ receptor, was substituted with NBD to provide the CB₂ receptor–selective NMP6 (Fig. 4; Table 3). Confocal microscopy demonstrated specific binding of NMP6 to CB₂ receptor on CD4⁺ T cells, as binding was inhibited by preincubation with the selective CB₂ receptor agonist GW842166X. Further utility of NMP6 was demonstrated using flow cytometry to study CB₂ receptor expression on mouse lung mononuclear B cells.

Use of commercially available T1117 (tocrifluor 1117) (Fig. 4; Table 3) was first reported in 2008 (Daly et al., 2008). T1117 was developed by linking AM251 [a CB₁ receptor–selective inverse agonist, IC₅₀ = 4 ± 1 nM (Gatley et al., 1997)] to a tetramethylrhodamine fluorophore. There are conflicting reports regarding pharmacological characterization—early studies reported T1117 to bind to GPR55 (increased Ca²⁺ response in HEK293 cells expressing GPR55) with only weak affinity at CB₁ receptor (Daly et al., 2010). However, in a different study, T1117 exhibited moderate binding at CB₁ receptor (K_d = 460 ± 80 nM at rCB1) and was used in a CB₁ receptor competition-binding assay (Bruno et al., 2014). The fluorescence of T1117 was quenched upon CB₁ receptor binding and restored upon displacement by unlabeled test compounds. T1117 was then used as the competitive tracer in a binding assay, providing IC₅₀ values for anandamide and AM251 in agreement with literature values. Bruno et al. (2014) further demonstrated utility of T1117 as a drug discovery tool by the appropriate identification of the CB₁ receptor allosteric ligand ORG27569 (Fig. 6). However, use of T1117 to study CB₁ receptor in native cell environments is limited, as it exhibits nonspecific binding to membrane most likely due to high lipophilicity (Bruno et al., 2014). This limitation is probe specific and not a general feature of fluorescent ligands, as demonstrated by fluorescent ligands for FFA1 receptor (discussed in this work) and for other class A GPCRs (reviewed in Vernall et al., 2014) that can be used to study GPCRs in native cell environments.

CB receptor agonists HU210 and HU308 were derivatized with a biotin tag at the ethoxy position, suitable for subsequent in situ conjugation to a fluorophore (Martín-Couce et al., 2012). The biotin-derivatized HU210-1 exhibited high affinity but little subtype selectivity for CB receptors, whereas biotin-derivatized HU308-3 exhibited selectivity for CB₂ receptor (Fig. 4; Table 3). Endogenous CB₁ and CB₂ receptor expression was studied in neurons and microglia using biotin-HU210-1 and biotin-HU308-3 by addition of streptavidin-Alexa488 fluorophore, and receptor-specific binding was then confirmed by using unlabeled HU210. Martín-Couce et al. (2012) also used biotin-HU210-1 in flow cytometry to study CB receptor expression in the monocytic cell line THP-1 at the single-cell level. The same research group has used biotin-HU210-1 in conjugation with streptavidin-Alexa488 to show there is high CB₁ receptor expression in B, T, plasmacytoid dendritic, and myeloid dendritic cells from donors with allergic rhinitis, atopic dermatitis, or food allergies (Martin-Fontecha et al., 2014).

Recently, within the space of a few months, there were three independent reports of fluorescent ligands for FFA1 receptor. In one of these reports, fluorescent ligands based on the TUG-770 or TUG-905 pharmacophore linked to a NBD fluorophore were constructed (Christiansen et al., 2016). Of these, TUG-905-NBD-4 (Fig. 5; Table 3) showed only a small reduction in potency compared with TUG-905 and retained FFA1 receptor agonism. A BRET assay paired with NanoLuciferase-tagged FFA1 receptor was established using TUG-905-NBD-4, with low levels of nonspecific fluorescence allowing for measurement of ligand kinetic parameters, and also a robust competition-binding assay was established. In a different study, a fluorescein-conjugated TAK-875 (Fig. 1) ligand (F-TAK-875A) (Fig. 5; Table 3) was synthesized as a racemic mixture and used in flow cytometry competition assays, which showed that FFA carbon-chain length was correlated with binding potency to FFA1 receptor (Ren et al., 2016). In the third report, Bertrand et al. (2016b) synthesized and evaluated a series of TAK-875-linker–fluorophore conjugates containing different types of linkers and fluorophores. In live cells overexpressing hFFA1 receptor, the lead probe TAK-875-Alexa488-16 (Fig. 5; Table 3) showed specific labeling of FFA1 receptor. This fluorescent agonist may prove especially valuable for studying dynamic receptor processes as 90-minute postincubation fluorescent internalization was observed. Endogenously expressed FFA1 receptor in pancreatic β cells was then visualized using a combination of TAK-875-Alexa488-16 and an Alexa488 antibody to amplify the signal. These reports collectively highlight the importance of both linker position and physicochemical properties for a successful fluorescent imaging tool. All three of these reports were based on a FFA1 receptor–selective parent pharmacophore (TAK-875 or TUG-905 core); however, the affinity and/or function of the fluorescent tool at the other FFA receptors were not provided, and it can be predicted, but not assumed, that the fluorescent tool and parent pharmacophore have a comparable receptor subtype selectivity profile.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

FFA and prostanoid receptor fluorescent tools.

To our knowledge, the only report of high-affinity synthetic ligand fluorescent conjugates for prostanoid, leukotriene, PAF, or GBPA receptors is by Tomasch et al. (2012), who synthesized a series of cinnamic acid antagonists containing various fluorophores. The most promising fluorescent compound when accounting for both EP₃ receptor affinity and a desirable emission wavelength above that of tissue autofluorescence was pyryllium-labeled cinnamon acid derivative 8 (Fig. 5; Table 3). Although with approximately threefold reduced affinity for EP₃ receptor compared with the parent, nonfluorescent pharmacophore, in different cell lines with respective recombinant EP_1–4 receptor subtypes, eight showed selectivity for EP₃ receptor over EP₁ (550-fold), EP₂ (24-fold), and EP₄ receptor (sevenfold). Labeling of EP₃ receptors in HT-29 cells using eight was demonstrated, with binding determined as specific by displacement of eight with excess nonfluorescent selective EP₃ receptor ligand. Further utility of eight was demonstrated in murine kidney, human brain tissue, and human platelets, with the authors commenting that future studies will aim to improve EP₃ over EP₄ receptor selectivity.

In the same body of work reporting photoactivatable ginkgolide derivatives as PAF receptor tools (Photoactivatable Covalent Tools), two ginkgolide derivatives tagged with a dansyl fluorophore at different positions were reported (Strømgaard et al., 2002). The most promising of these fluorescent conjugates had moderate PAF receptor affinity [Ki = 0.96 µM, guinea pig (gp) PAF]; however, it was not evaluated further for PAF receptor fluorescent labeling or photoactivated covalent binding. Nevertheless, future development of fluorescently tagged ginkgolide derivatives as fluorescent tools may prove useful for PAF receptor interrogation.

2. Near-Infrared Wavelength Fluorescent Tools for In Vivo Imaging.

Both the optical properties of tissues, including endogenous chemical components present, and choice of fluorophore can influence the depth that a suitable signal-to-noise fluorescent signal can be detected in vivo (reviewed in Pansare et al., 2012; Hong et al., 2017). Longer wavelength fluorophores such as those in the near-infrared (NIR) region (700–900 nm) and the now termed NIR-II region (1000–1700 nm) are well suited to in vivo imaging and typically can be detected at a tissue depth of up to 5–7 mm with good signal-to-noise resolution. As is the case with in vivo radioligands, the absorption, distribution, metabolism, and excretion of the tool in relation to the intended use need to be considered, along with careful characterization of in vivo versus in vitro selectivity.

There has been particular interest surrounding in vitro use/characterization and also in vivo use of NIR fluorescent tools to study CB₂ receptor expression in malignant tumors. Synthesis of selective CB₂ receptor SR144528 derivatives with linkers introduced to the pyrazole or 4-chloro phenyl position abolished CB₂ receptor–binding affinity; however, incorporation of a linker at the benzylic position was more successful, resulting in the SR144528 linker derivative mbc94 (K_i = 15 nM at mCB₂) (Fig. 4) (Bai et al., 2008; Sexton et al., 2011). Attachment of the fluorescent IRDye 800CW to mbc94 to give NIR-mbc94 (Fig. 4; Table 3) led to some loss in affinity (K_i = 260 nM at mCB₂). This fluorescent tool was successfully used in a multiwell high throughput screening assay using intact CB₂-mid delayed brain tumor (DBT) cells (mouse DBT cells transfected with CB₂ receptor); however, imaging experiments using primary microglia cells revealed a high level of NIR-mbc94–nonspecific binding (Sexton et al., 2011).

The same group of researchers developed NIR760-mbc94 (Fig. 4; Table 3) using the same CB₂ receptor pharmacophore-linker mbc94 but a different fluorophore (NIR760) (Zhang et al., 2013). When CB₂-mid DBT cells were incubated with NIR760-mbc94 in the presence and absence of the control nonfluorescent pharmacophore SR145528, the fluorescent intensity was reduced by only 40% compared with the control, indicating some nonspecific binding of the fluorescent tool, which the authors suggested was due to nonspecific protein binding because NIR760-mbc94 has a net negative charge. In a cancer model, CB₂-mid DBT cells were injected subcutaneously into the right flank of healthy mice, and tumors were allowed to develop for 10 days. Zhang et al. (2015) then injected mice with NIR760-mbc94 and observed fluorescence on the whole body, followed by tumor-specific localization 48–72 hours postinjection. Injection with SR144528 1 hour prior to NIR760-mbc94 was carried out to determine specific CB₂ receptor binding of NIR760-mbc94, and after 72 hours a 31% reduction in fluorescence of the tumor area/normal area ratio was observed. As CB₂ receptor plays an important role in inflammation, this fluorescent tool has also been used for in vivo imaging of mCB₂ receptor expression in a complete Freund’s adjuvant–induced inflammation mouse model (Zhang et al., 2015). Low specific binding of NIR760-mbc94 was observed with inflammation-specific fluorescence visible only 36 hours postinjection.

The same group of researchers then conjugated a quinolone-based ligand with the NIR760 fluorophore to give NIR760-Q (Fig. 4; Table 3) (Wu et al., 2014b). Nonspecific binding was observed in experiments with Jurkat cells in the presence of the unlabeled control ligand 4Q3C, which the authors suggest may be due to the total negative charge of NIR760. The same research group subsequently developed the zwitterionic NIR-fluorophore–containing ZW760-mbc94 (Fig. 4; Table 3) (Wu et al., 2014a). Specificity of ZW760-mbc94 for CB₂ receptor was determined in vitro by incubating CB₂-mid DBT cells in the absence or presence of the unlabeled control ligand 4Q3C, which afforded a 50% reduction in fluorescence intensity. This is a moderate improvement compared with negatively charged NIR760-mbc94, and development of fluorescent tools with improved in vivo selective binding is an ongoing challenge. The fluorescent tool ZW760-mbc94 was further evaluated by injection into mice 10 days after CB₂-mid DBT cell inoculation in the right flank, and images showed high fluorescence in the liver as well as fluorescence throughout the whole body that persisted 72 hours after ZW760-mbc94 injection. Ex vivo analysis revealed mice treated with the control blocking ligand 4Q3C showed a 47% reduced tumor:normal tissue fluorescence ratio ascompared to ZW760-mbc94–treated mice without blocking ligand, indicating some observed fluorescence was CB₂ receptor-dependent.

Another fluorescent tool developed by the same research group is NIR760-XLP6 (Fig. 4; Table 3), which consists of a pyrazolopyrimidine pharmacophore conjugated via a linker to the NIR760 fluorophore (Ling et al., 2015). Experiments were carried out to determine in vivo receptor selectivity, which showed higher fluorescence (40% higher) in mice with CB₂-mid DBT tumors as compared with CB₁-mid DBT (DBT cells transfected with CB₁ receptor) tumors. This research group has also developed a CB₂ receptor–targeted photosensitizer IR700DX-mbc94 (Fig. 4; Table 3) by attaching a NIR fluorophore, IR700DX, to mcb94 (Zhang et al., 2014). In vivo experiments in mice revealed that IR700DX-mbc94 inhibited growth of CB₂ receptor–positive tumors following light (wavelength 670–710 nm) irradiation, but not that of the CB₂ receptor-negative tumors (Jia et al., 2014). This is an interesting novel approach to treating tumors overexpressing CB₂ receptor. A patent application (Bornhop et al., 2013) has been published by the same group of researchers that describes the use of SR144528 derivatives tethered to an IRDye800CW fluorophore, a topoisomerase inhibitor (such as etoposide), and a gadolinium chelate. The potential applications of these SR144528 derivatives claimed by the patent include use as a molecular imaging tool and as a targeted drug delivery system.

The development and use of fluorescent tools for GPCRs are a rapidly developing field. There are several reports of selective CB₂ receptor fluorescent tools; however, there is a lack of selective CB₁ receptor fluorescent tools, and no doubt this is an area of intensive research. Selective fluorescent tools for the other FFA, S1P, LPA, prostanoid, leukotriene, PAF, and GPBA receptors based on synthetic pharmacophores will most likely be developed in the future, which will facilitate some exciting biologic experiments.

1. Photoactivatable Covalent Tools.

The most commonly used inert groups capable of photoactivation are azides, benzophenones, and diazirines. An advantage of photoactivated ligands is that the highly reactive species generated in situ is able to form a covalent bond with any amino acid nearby, compared with electrophilic covalent ligands that are limited to reaction with only one or two nucleophilic amino acid side chains (Sumranjit and Chung, 2013). The highly reactive species generated can covalently bind to the receptors through insertion into bonds (e.g., C-H) in the peptide backbone and amino acid side chains (Cavalla and Neff, 1985). Desirable properties of a photoactivated ligand include high receptor affinity, activatable at a wavelength that does not cause damage to the target receptor or biosystem, the generation of a reactive group with a short lifetime, indiscriminate or finely-tuned formation of a covalent bond with any nearby amino acid, and formation of a stable irreversible adduct (Vodovozova, 2007).

Initial efforts to develop covalent tools for CB receptors focused on analogs of ∆⁸-THC, which has a similar affinity for CB₁ receptor, but is more stable than the naturally more prevalent ∆⁹-THC isomer. A photoactivatable azido group was positioned at the THC aliphatic chain terminus, as SAR showed this position to be tolerant of bulky groups (Charalambous et al., 1992). This ligand, 5′azido-∆⁸-THC AM91 (Fig. 6; Table 4), exhibited approximately twofold increased affinity to that of ∆⁸-THC (K_i = 35 ± 11 nM at rCB₁). AM91 was developed shortly after cloning of CB₁ receptor with the intention of aiding receptor isolation and characterization and was a prototype for the next generation of photoactivatable ligands. Increasing interest in CB receptors and the need for tools to identify receptor subtypes and enable receptor isolation, purification, and characterization drove the development of subsequent CB receptor covalent tools.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

CB receptor photoactivatable, electrophilic, and bifunctional covalent tools.

The same research group radioiodinated AM91 by introducing [¹²⁵I] ortho to the phenolic group to give 2-iodo-5′azido-∆⁸-THC (2-[¹²⁵I]-AM91) (Fig. 6; Table 4) (Burstein et al., 1991). The 2-[¹²⁵I]-AM91, equilibrated with tissue and exposed to UV light to induce azide to nitrene conversion, could detect CB receptor expression in mouse cerebral cortex (K_d = 5.60 pM) and mouse lymphoma cells (K_d = 9.38 pM). Samples of mouse brain analyzed by SDS-PAGE and autoradiography showed formation of a covalent bond between ligand and CB receptor. This experiment provided the earliest evidence of a covalent ligand–CB receptor entity and provided a clue to the existence of receptor subtypes as a fainter, lower molecular weight band on the gel was also observed. This other band was later shown to be CB₂ receptor (Makriyannis, 2014).

Distinct chemical classes of photoaffinity labels can provide insight into the ligand-binding mode and the topography of the CB receptor ligand binding site(s). A series of photoactivatable CB receptor ligands featuring a heteroaroyl group at C3 in place of an alkyl chain were synthesized, including a CB₂ receptor–selective 3-benzothiophenyl-derivative AM967 (K_i = 34.2 nM at mCB₂, 124.8 nM at hCB₂, 1254 nM at rCB₁) (Fig. 6; Table 4) (Dixon et al., 2012). Photolysis of aryl phenones with UV light at approximately 350 nm generates a highly reactive triplet-state ketone intermediate, which can form covalent bonds through insertion in C-H bonds (Vodovozova, 2007). Receptor photolabeling of CB₂ receptor using AM967 was measured at 67% (reduction in specific binding of [³H]CP-55,940) in HEK293 membrane preparations expressing mCB₂ receptor (Dixon et al., 2012). This covalent series provided interesting SAR data on heteroaroyl-containing CB₂ receptor ligands. Future experiments with AM967 could elucidate key amino acids specifically involved in the binding of arylphenone analogs, expanding the structural understanding of CB₂ receptor and further enabling structure-based design of subtype-selective drugs.

The classic CB-based tool AM993, which has a C3 adamantyl group, has been synthesized to inform receptor molecular recognition of this fixed conformation adamantyl moiety (Ogawa et al., 2015). AM993 (Fig. 6; Table 4) behaved as an agonist at rCB₁ receptor with an EC₅₀ = 2.4 nM and E_max = 45% and as a neutral antagonist at hCB₂ receptor. Successful covalent labeling of CB receptors following pretreatment with 10-fold above the K_i of AM993 was demonstrated with a 67% reduction in specific binding of [³H]CP-55,940 at rCB₁ and a 60% reduction at hCB₂ receptor. A series of electrophilic and photoactivatable aryl pyrazole compounds based on the high-affinity CB₁ receptor antagonist rimonabant have been developed (Howlett et al., 2000). However, most of these compounds did not maintain the affinity of the parent compound, and the best in this series irreversibly labeled rCB₁ receptor at IC₅₀ = 28 nM. Modified ECBs have also been developed as CB receptor covalent ligands. AM3661 is based on the structure of AEA and incorporates a cyclopropylamide moiety at the head group for improved CB₁ affinity and a photoactivatable azide at the terminal alkyl chain. AM3661 (Fig. 6; Table 4) was shown to irreversibly label 68% of rCB₁ receptor when used at 18 nM (Li et al., 2005). Binding assays were carried out in the presence of a serine protease inhibitor to avoid AM3661 hydrolysis by fatty acid amide hydrolase (FAAH), a key enzyme in ECB metabolism. ECB-based covalent tools can be used to reveal key information regarding the binding and signaling of AEA and 2-AG as well as ECB-like ligands.

There are also a number of CB receptor photoactivatable tools reported in review literature but without published primary experimental data, which are summarized below. The tool 7′-N₃-1′,1′-dimethylheptyl-∆⁸-THC, the 1′-geminal dimethyl and seven carbon aliphatic chain analog of AM91, was reported to give improved affinity (K_i = 0.4 nM at CB₁) compared with AM91 (Picone et al., 2002). This improved affinity would be advantageous when working in native cell lines with low CB receptor expression. The (-)-11-hydroxy-7′azido-∆⁸-THC (AM836) was ascribed an IC₅₀ = 0.16 nM at CB₁ receptor (Palmer et al., 2002; Thakur et al., 2005). Two tools containing an azido instead of the 10-methyl of 1′,1′-dimethylheptylhexahydrocannabinol have been reported. AM869 contains an iodide at the terminus of the seven-carbon chain (K_i = 0.67 nM at CB₁, 0.72 nM at CB₂), whereas AM1708 has both an alkene and ¹²⁵I at the terminus of the alkyl chain (K_i = 0.8 nM at CB₁, 0.85 nM at CB₂) (Khanolkar et al., 2000).

Initial characterization of CysLT₁ receptor was aided by a radioiodinated photoactivatable azido analog of LTD₄. [¹²⁵I]-azido-LTD₄ [K_i = 1.7 nM (nonradioactive analog)] (Fig. 7; Table 4) was able to selectively covalently label a 45-kDa protein in guinea pig lung membranes (Metters and Zamboni, 1993). LTD₄, LTE₄, LTC₄, and MK-0571 (a CysLT₁ antagonist) were able to inhibit the photolabeling of the 45-kDa protein by [¹²⁵I]-azido-LTD₄ with similar potencies to their IC₅₀ values at CysLT₁ receptor, providing evidence that this protein is CysLT₁ receptor. However, [¹²⁵I]-azido-LTD₄ was also found to nonselectively label guinea pig serum albumin. Subsequently, the more synthetically accessible photoaffinity probe [¹²⁵I]-L-745310, an analog of CysLT₁ receptor antagonist montelukast, containing a trifluoromethyl diazirine as the photoactivatable group, was developed (Fig. 7; Table 4) (Gallant et al., 1998). The reactive carbene generated upon protolysis (350 nm) of the diazirine can insert into C-H bonds. [¹²⁵I]-L-745310 was able to label the same 45-kDa protein in guinea pig lung preparations, acting as an antagonist (IC₅₀ = 27 nM, 53 nM at gpCysLT₁). High levels of nonspecific labeling were reduced with addition of a detergent and optimization of irradiation time and temperature.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

Leukotriene, PAF, and prostanoid receptor photoactivatable covalent tools.

The endogenous leukotrienes possess some innate photoactivatable labeling properties due to the presence of conjugated double bonds, and this has been exploited (Falk et al., 1989; Müller et al., 1991; Slipetz et al., 1993; Nicosia et al., 1995). However, this process requires cryofixation of the receptor–ligand complex to reduce significant nonspecific labeling.

Photoactivatable peptidoleukotriene analogs for CysLT₂ receptor have been developed, containing either 7Z,9E or 7E,9E dienic moieties in an effort to improve stability of the endogenous leukotrienes (Klotz et al., 1993). The aryldiazonium-substituted dienic LTD₄ derivatives (Fig. 7; Table 4) showed greater selectivity for CysLT₂ receptor than the less potent arylazido derivatives. In addition, irradiation of the azido analogs at 245 nm overlaps with the absorption of the dienic moiety (235 nm), resulting in simultaneous decomposition of both azide and diene, whereas only a single reactive species is produced when the diazonium analogs are irradiated at 360 nm.

Characterization of PAF receptor was aided by a similar approach using [¹²⁵I]AAGP, a photoactivatable, radiolabeled derivative of PAF (Fig. 7; Table 4) (Chau et al., 1989). [¹²⁵I]AAGP showed saturable, high-affinity binding in rabbit platelet membranes (K_d = 2.4 ± 0.7 nM, B_max = 1.1 ± 0.2 pmol/mg; EC₅₀ of nonradioactive analog = 3.2 ± 1.9 nM), and photolabeling identified a protein of 52 kDa as the PAF receptor. In another study, a series of ginkgolide B and ginkgolide C derivatives were synthesized containing varying photoactivatable groups (Strømgaard et al., 2002). These were all found to act as antagonists at PAF receptor with the highest affinity compound a ginkgolide B derivative containing a tetrafluorophenylazide (K_i = 90 nM at gpPAF) (Fig. 7; Table 4) and being more potent than ginkgolide B (K_i = 0.56 µM at gpPAF receptor). This ligand could potentially be used to investigate ginkgolide interactions with PAF receptor; however, as yet covalent binding has not been demonstrated.

The first prostanoid receptor covalent tool developed was a diazonium salt of 9,11-dimethylmethano-11,12-methano-16-(4-aminophenoxy)13,14-dihydro-13-aza-15 alpha beta-omega-tetranor TXA2, PTA-POA, which was shown to irreversibly bind to hTP receptor, but affinity was unable to be determined (Mais et al., 1986). An iodinated azide derivative of 13-azaprostanoic acid, I-APA-PhN₃ (Fig. 7; Table 4), was able to irreversibly bind TP receptor with moderate affinity (K_i = 290 nM at hTP) and irreversibly inhibit human platelet aggregation (Arora et al., 1987; Kattelman et al., 1987). The TP receptor photoaffinity probe, [¹²⁵I]PTA-azido (Fig. 7; Table 4), had good affinity (K_d = 11 nM at hTP) and was able to label three protein bands of 43, 39, and 27 kDa, but it was not clear which of these was TP receptor (Mais et al., 1989). [¹²⁵I]PTA-azido was used to determine the isoelectric point of TP receptor (Mais and Halushka, 1989).

Another TP receptor photoaffinity probe, I-PTA-PON₃ (Fig. 7; Table 4), was used to inhibit human platelet aggregation and irreversibly label hTP receptor (Mais et al., 1990). A radioiodinated photoaffinity probe [¹²⁵I]SAP-N₃ (Fig. 7; Table 4) irreversibly labeled human platelet TP receptor with high affinity (Mais et al., 1991). [¹²⁵I]SAP-N₃ was used with SDS-PAGE autoradiography to identify TP receptor with a molecular weight of 50–51 kDa, and subsequent digestion of this photoaffinity-labeled receptor identified two N-linked glycosylation sites (Mais et al., 1992). Proteolytic cleavage studies of purified, [¹²⁵I]SAP-N₃–labeled TP receptor were carried out to localize the ligand binding domain to amino acids 99–192 (True and Mais, 1994).

A PGI₂ analog, [³H]APNIC (Fig. 7; Table 4), was used to characterize IP receptor. [³H]APNIC had good affinity (K_d = 4.7 nM, B_max 0.58 pmol/mg protein, mouse mastocytoma P-815 cells) and photolabeled IP receptor with good efficiency (80% of specific binding of [³H]APNIC at 13 nM), and was able to identify a 43-kDa protein in mouse mastocytoma and a 45-kDa protein in porcine platelets as IP receptor (Ito et al., 1992; Suzuki et al., 1992).

A moderate affinity azidophenacyl ester of PGE₂ (azido-PGE₂) (Fig. 7; Table 4) was synthesized for use as a tool to isolate and identify EP receptor (Michalak et al., 1990). The radiolabeled analog, [³H]azido-PGE₂, was able to covalently label a protein of 100 kDa in isolated bovine cardiac sarcolemmal vesicles, and this photolabeling could be inhibited with excess unlabeled PGE₂ and azido-PGE₂. Another EP receptor photoaffinity probe has been reported, (15S)-17-(4-azidophenyl)-18,19,20-trinorprostaglandin E2, which had moderate receptor affinity (IC₅₀ = 300 nM chicken spinal cord EP receptor), but covalent binding was not demonstrated (Kawada et al., 1991). A series of photoactivatable PGF_2α derivatives were synthesized with the purpose of developing an FP receptor probe. The highest affinity derivative in the series had a K_i of 49 nM (ovine luteal cells); however, covalent binding wasn’t demonstrated (Golinski et al., 1992).

A number of other photoactivatable ligands have been developed for lipid-binding receptors that, due to low-moderate affinity, will have limited utility as covalent tools, but may still have some application. For instance, labeling with moderate affinity aryl azide LTB₄ derivative photoaffinity probe 4bα (IC₅₀ = 0.7 µM at hLTB4₁) (Fig. 7; Table 4) was carried out in conjunction with amino acid sequencing and mass spectrometry to determine the binding residues as Cys 97, Ser 100, Met 101, and Ser 104 of TM-III and Trp 234 and Tyr 237 in TM-VI, providing initial characterization of the LTB4₁ receptor ligand binding site (Durand et al., 2000).

2. Electrophilic Covalent Ligands.

A variety of reactive moieties has been used in covalent labels for class A GPCRs, including isothiocyanates, halomethylketones, reactive thiols, Michael acceptors, and nitrogen mustards (Weichert and Gmeiner, 2015). The isothiocyanate is a popular choice for electrophilic tools as it is readily prepared from primary amines (Wong and Dolman, 2007), is stable in water, but is reactive with amino acids containing a thiol, imidazole, or amine under physiologic conditions (Guo et al., 1994).

AM708 and the closely related analog 7′-NCS-DMH-THC were the first two high-affinity electrophilic covalent ligands reported for CB₁ receptor (Fig. 6; Table 4). Affinity of AM708 (IC₅₀ = 1.6 ± 0.3 nM at rCB₁) was measured with a [³H]CP-55,940 displacement assay (Guo et al., 1994). Treatment with 10 nM AM708 led to 80% reduction in the available rCB₁ receptor binding sites, whereas 100 nM treatment practically depleted all the rCB₁ receptor binding sites. Time- and AM708 concentration-dependent irreversible ligand–CB receptor binding was demonstrated for the first time. Guo et al. (1994) postulated that there must be an amine, thiol, or imidazole amino acid side chain present in the CB receptor binding site to which the covalent ligand bound. The more hydrophobic 7′-NCS-DMH-THC (IC₅₀ = 0.66 nM at rCB₁) demonstrated irreversible binding with incubation at five times the apparent IC₅₀ leading to an 83% reduction in specific binding of [³H]CP-55,940 at rCB₁ receptor (Morse et al., 1995). Taking all evidence into account, it was postulated that covalent binding of AM708 and 7′-NCS-DMH-THC indicates the likelihood of a lysine or cysteine residue in proximity to the alkyl side chain terminus in the CB₁ receptor–binding pocket (Picone et al., 2002). The CB receptor ligand AM960 (Fig. 6) [IC₅₀ = 25 nM at rCB₁ (preliminary biologic experiment, no further full experiment reported)] was designed in an effort to introduce an iodo group while maintaining the covalent binding ability of AM708 (Chu et al., 2003). Incorporation of a C6 3-iodopropyne group was informed by previous SAR showing that an iodopropyl group at this position conferred high affinity and the iodo group offers potential for [¹²⁵I] radiolabeling.

Concurrent use of covalent ligands and site-directed mutagenesis is a powerful way to study ligand-binding orientation and determine which amino acids are involved in binding of a specific ligand. AM841 (Fig. 6; Table 4) was the first CB₁ receptor covalent tool to be used in conjunction with site-directed mutagenesis and provided important information about ligand activation sites (Picone et al., 2005). It was proposed using CB₁ receptor homology modeling and ligand docking that a cysteine residue present on helix 6 C6.47 (residue 355), part of the highly conserved CWXP-binding motif in class A GPCRs, was the most likely site of covalent attachment of AM841. Mutation of Cys C6.47 to Ser, Ala, or Leu to reduce or eliminate the nucleophilicity of residue 355 was carried out to test this hypothesis. The affinity of AM841 for hCB₁ C6.47 Ser (K_i = 10.46 ± 0.88 nM) and Ala (K_i = 11.32 ± 0.29 nM) mutants was similar but was reduced for the C6.47 Leu mutant (K_i = 58.09 ± 11.69 nM). In spite of retaining some receptor affinity, AM841 did not irreversibly bind to any of the three mutants. It is therefore worth considering the contribution of measured binding affinity versus covalent bond formation—high ligand affinity is not dependent on covalent bond formation, although measurement can be biased depending on the experiment used to measure K_i. Also of consideration is the effect site-directed mutagenesis has on overall receptor shape/misfolding. In this study, the CB₁ receptor global conformation of the mutants is likely to be maintained as it is in the wild-type receptor because AM841 displayed comparable affinity.

There have been multiple studies utilizing AM841 to probe CB receptor structure and function. Pei et al. (2008) used AM841 to demonstrate distinctions in ligand-binding motifs between hCB₁ and hCB₂ receptor subtypes via complementary site-directed mutagenesis and ligand-docking studies. Mutagenesis of hCB₂ receptor indicated that C6.47 is the site of covalent attachment of AM841 and ruled out two cysteines on helix seven, C7.38 and C7.42. Activation of hCB₂ receptor by AM841 (leading to inhibition of forskolin-stimulated cAMP production) was achieved with much higher potency compared with the noncovalent analog of AM841 (wherein the NCS group is replaced by an H atom) or at hCB₁ receptor. The highly conserved CB₁ receptor Lys K3.28, which has been shown to be important in recognition of multiple ligands by hCB₁ receptor (Song and Bonner, 1996), was demonstrated as having little effect on hCB₂ receptor–AM841 binding. Pei et al. (2008) also postulated that AM841 accesses the CB₂ receptor–binding pocket through the lipid bilayer (discussed in Leukotriene Receptor). Despite a potential lipid bilayer entry, covalent ligands such as AM841 are thought unlikely to bind to random nucleophilic amino acids as a high-affinity binding interaction needs to occur before significant covalent labeling can occur. This specific covalent reactivity can be shown using mass spectrometry, as demonstrated by Szymanski et al. (2011), who further characterized the covalent interaction of AM841 with hCB₂ receptor residue C6.47. Multiple reaction mass spectrometry monitoring showed that covalent modification of hCB₂ receptor by AM841 was exclusive to TMH6, and high-resolution mass spectrometry of the TMH6 tryptic peptide confirmed this covalent labeling to be selective for C6.47 (Szymanski et al., 2011). This report shows the power of combining mass spectrometry–based proteomics and site-directed covalent labeling in the elucidation of GPCR ligand binding sites. Taken together, this information provides valuable insight into CB receptor ligand-binding pocket(s) and requirements for high-affinity, selective ligand design.

Highly specific and potent covalent ligands have also been used to investigate the physiologic and pathophysiological roles of GPCRs. AM841 has been used in an in vivo study of inflammatory bowel disease to show the involvement of both central and peripheral mCB₁ and mCB₂ receptors in the anti-inflammatory action of cannabinoids (Fichna et al., 2014). The utility of AM841 in research and therapeutics has been further demonstrated by examining gastrointestinal motility in healthy and stressed mice (Keenan et al., 2015). In this study, AM841 was found to act as a peripherally restricted ligand, normalizing accelerated gastrointestinal motility through action on CB₁ receptor in the small and large intestine. These studies demonstrate the therapeutic potential of covalent ligands such as AM841. It would be interesting to see what the equivalent noncovalent ligand would do in the same experiments to dissect the importance of the covalent attachment in the observed therapeutic effects.

An electrophilic adamantyl-substituted covalent tool (AM994) (Fig. 6; Table 4) has also been developed with good affinity for both CB₁ and CB₂ receptors, which behaved as an agonist at rCB₁ receptor and an inverse agonist at hCB₂ receptor (Ogawa et al., 2015). Successful labeling of CB receptor following pretreatment with 10-fold the K_i of AM994 was demonstrated with a 63% reduction in specific binding of [³H]CP-55,940 at rCB₁ receptor and a 74% reduction at hCB₂ receptor.

The same research group that developed photoactivatable AM3661 has also synthesized an isothiocyanate-functionalized AEA tool AM3677 (Fig. 6; Table 4), which irreversibly labels 58% of rCB₁ receptor when used at 26 nM (binding assays carried out in presence of serine protease inhibitor as with AM3661) (Li et al., 2005). Further structural and functional profiling was carried out by Janero et al. (2015), which showed that AM3677 forms a covalent bond to hCB₁ C6.47. AM3677 was found to function as an agonist, inhibiting cellular cAMP formation and stimulating irreversible internalization of rCB₁ receptor. The authors propose AM3677 could be used as a tool in the study of ECB-induced CB₁ receptor activation and associated signaling.

Functionalization of a known biarylpyrazole (AM6731) with an isothiocyanate group generated AM1336 (Fig. 6; Table 4), a covalent hCB₂ receptor inverse agonist (Mercier et al., 2010). AM1336 irreversibly bound 60% of available hCB₂ receptor when administered at 5.4 nM. Mutation of Cys to Ala or Ser at single and multiple points in hCB₂ receptor and analysis using tool AM1336 revealed two residues in TMH7 (C7.38 residue 284 and C7.42 residue 288) as key for inverse agonist binding. C1.39 (residue 40) of TMH1 was also found to modulate hCB₂ receptor ligand affinity, and residue C137 from intercellular loop 2 was shown to affect the maximum efficacy of AM1336.

Exploration of the aminoalkylindole CB chemical scaffold as an electrophilic ligand has been carried out with synthesis of analogs containing an isothiocyanate on the indole ring or at various C3 naphthyl positions (Yamada et al., 1996). The most potent derivative (isothiocyanate 12) (Fig. 6; Table 4), containing indole 6-isothiocyanate substitution, demonstrated irreversible binding at sixfold the IC₅₀ with 70% reduction in specific binding of [³H]CP-55,940 at rCB₁ receptor. This depletion of [³H]CP-55,940 binding indicates that the aminoalkylindole class and nonclassic CBs (e.g., CP-55,940) may have overlapping binding sites.

Due to the psychotropic side effects associated with direct CB₁ receptor activation, alternative CB drug development strategies have been investigated. One area of promise is allosteric modulators, potentially allowing more efficacious control of downstream signaling effects. However, improved knowledge of any CB₁ receptor allosteric site(s) is critical to drug development. To this end, Kulkarni et al. (2016) have recently developed the covalent CB₁ receptor ligand GAT100 (Fig. 6; Table 4), by replacement of the chloro moiety of allosteric CB₁ receptor allosteric ligand ORG27569 (Fig. 6) with an isothiocyanate group. Further in-depth characterization showed GAT100 is a negative allosteric modulator of CP-55,940 (Fig. 1), AEA, and 2-AG across several signaling pathways and may interact with key residue C7.38(382) of CB₁ receptor (Laprairie et al., 2016). It will be very interesting to see what structural knowledge of the CB₁ receptor allosteric site is gleaned by future use of this tool.

Methyl arachidonyl fluorophosphonate has been shown to act as an irreversible inhibitor of CB₁ receptor (IC₅₀ = 20 nM at rCB₁) (Fig. 6; Table 4), preventing subsequent binding of CP-55,940 and reducing the maximal responses of agonists WIN 55,212-2 and CP-55,940 (Deutsch et al., 1997; Fernando and Pertwee, 1997). Methyl arachidonyl fluorophosphonate, which is also a potent irreversible antagonist of FAAH (IC₅₀ = 2.5 nM at rFAAH), acts as a phosphonylation agent as the electrophilic fluorophosphonate group can covalently label nucleophilic residues such as serine.

There is a lack of electrophilic covalent probes for the other fatty acid–binding class A GPCRs when compared with CB receptors. An attempt at an electrophilic covalent ligand for the putative prostamide receptor(s) was made; however, biologic data have not yet been published (Shelnut et al., 2015).

3. Bifunctional Covalent Ligands.

Bifunctional ligands with two reactive/photoactivatable functional groups are capable of forming two covalent bonds to one receptor. To this end, the homobifunctional ligand AM5823 with two isothiocyanate groups and the heterobifunctional ligand AM5822 with an isothiocyanate and an azide group have been developed (Fig. 7) (unpublished data) (Makriyannis, 2014). Another bifunctional tool, di-azido AM859, has been reported in a review (K_i = 1.6 nM at CB₁, 2.65 nM at CB₂) (unpublished primary experimental data) (Khanolkar et al., 2000). It is hoped that this approach will provide greater spatial accuracy in the classification of CB receptor ligand–binding orientation.