**Radiosynthesis and Biological Investigation of a Novel Fluorine-18 Labeled Benzoimidazotriazine-Based Radioligand for the Imaging of Phosphodiesterase 2A with Positron Emission Tomography**

**Rien Ritawidya 1,2,\*, Barbara Wenzel 1, Rodrigo Teodoro 1, Magali Toussaint 1, Mathias Kranz 3,4, Winnie Deuther-Conrad 1, Sladjana Dukic-Stefanovic 1, Friedrich-Alexander Ludwig 1, Matthias Scheunemann <sup>1</sup> and Peter Brust <sup>1</sup>**


Received: 24 October 2019; Accepted: 12 November 2019; Published: 15 November 2019

**Abstract:** A specific radioligand for the imaging of cyclic nucleotide phosphodiesterase 2A (PDE2A) via positron emission tomography (PET) would be helpful for research on the physiology and disease-related changes in the expression of this enzyme in the brain. In this report, the radiosynthesis of a novel PDE2A radioligand and the subsequent biological evaluation were described. Our prospective compound 1-(2-chloro-5-methoxy phenyl)-8-(2-fluoropyridin-4-yl)-3 methylbenzo[e]imidazo[5,1-c][1,2,4]triazine, benzoimidazotriazine (**BIT1**) (IC50 PDE2A = 3.33 nM; 16-fold selectivity over PDE10A) was fluorine-18 labeled via aromatic nucleophilic substitution of the corresponding nitro precursor using the K[18F]F-K2.2.2-carbonate complex system. The new radioligand [18F]**BIT1** was obtained with a high radiochemical yield (54 <sup>±</sup> 2%, n = 3), a high radiochemical purity (≥99%), and high molar activities (155–175 GBq/μmol, n = 3). In vitro autoradiography on pig brain cryosections exhibited a heterogeneous spatial distribution of [18F]**BIT1** corresponding to the known pattern of expression of PDE2A. The investigation of in vivo metabolism of [18F]**BIT1** in a mouse revealed sufficient metabolic stability. PET studies in mouse exhibited a moderate brain uptake of [18F]**BIT1** with a maximum standardized uptake value of ~0.7 at 5 min p.i. However, in vivo blocking studies revealed a non-target specific binding of [18F]**BIT1**. Therefore, further structural modifications are needed to improve target selectivity.

**Keywords:** cyclic nucleotide phosphodiesterase; PDE2A radioligand; nitro-precursor; fluorine-18; in vitro autoradiography; PET imaging

#### **1. Introduction**

The cyclic nucleotide phosphodiesterases (PDEs) are a superfamily of enzymes catalyzing the hydrolysis of the intracellular secondary messengers, cyclic adenosine monophosphate (cAMP) and

cyclic guanosine monophosphate (cGMP) [1–3]. These secondary messengers are involved in a great variety of cellular functions associated with normal and pathophysiological processes in the brain and periphery [3–6].

The 11 family members of PDEs are encoded by 21 genes and classified by their substrate specificity [3,7]. PDE 4, 7, and 8 are only cAMP-specific, PDE 5, 6, and 9 are cGMP-specific, and the remaining PDE 1, 2, 3, 10, and 11 hydrolyze both substrates [3,6,7].

The dual-substrate enzyme PDE2A is abundantly expressed in brain, in particular, in caudate, nucleus accumbens, cortex, hippocampus [3,8,9], amygdala [10,11], substantia nigra, as well as olfactory tubercle [11,12], while the expression in the midbrain, hindbrain, and cerebellum is comparatively low [8,11,12]. This specific distribution indicates a role of PDE2A in the modulation of complex neuronal processes, such as learning, concentration, memory, emotion, depression, anxiety, and CNS related disorder [10,13,14]. The pharmacological inhibition of PDE2A has been evaluated in preclinical studies, thus suggesting PDE2A inhibitors as a potential treatment for neurodegenerative diseases, such as Alzheimer's disease, schizophrenia, and dementia [15,16]. PDE2A inhibition has the potential to prolong the duration of cAMP- and cGMP-dependent signaling pathways, eventually improving neural plasticity and memory function [3,13,17].

Positron emission tomography (PET) is a molecular imaging modality that enables the visualization, characterization, and measurement of molecular targets and biochemical processes in living systems [18]. Accordingly, a PDE2A specific radiotracer would allow quantification of PDE2A expression, as well as disease-related changes thereof.

The so far most developed PDE2A radioligands are shown in Figure 1. The first highly potent PDE2A compound [18F]**B-23** (IC50 hPDE2A = 1 nM; IC50 rPDE10A = 11 nM) was developed by Janssen Pharmaceutica NV (Beerse, Belgium) [10,19]. Biodistribution and microPET imaging studies in rats demonstrated high uptake of activity in the striatum; however, brain-penetrating radio-metabolites limited further evaluation. Pfizer Inc. (New York, NY, USA) also reported on a highly affine PDE2A radioligand, [18F]**PF-05270430** (IC50 hPDE2A = 0.5 nM; IC50 hPDE10A >3000 nM) [10,20], which has been evaluated in monkeys [20] and translated to clinical trials [21,22]. [18F]**PF-05270430** showed high target-specific accumulation in the putamen, caudate, and nucleus-accumbens, as well as good metabolic stability and a favorable kinetic profile, pointing out [18F]**PF-05270430** as a promising PDE2A PET ligand [21]. However, using the cerebellum as a reference region, the estimated binding potential of [18F]**PF-05270430** was low, and the authors concluded that further studies are required to validate the suitability of the cerebellum as a reference region [21].

In parallel to Pfizer, our group developed further PDE2A radioligands on the basis of pyridoimidazotriazine. The recently published compounds [18F]**TA3** (IC50 PDE2A = 11.4 nM; IC50 PDE10A = 318 nM) and [18F]**TA4** (IC50 PDE2A = 7.3 nM; IC50 PDE10A = 913 nM) (Figure 1) were characterized by high potency and selectivity towards PDE2A [24]. Notably, both radiotracers were found to be suitable radioligands for in vitro imaging of PDE2A; however, in vivo metabolism studies revealed a high fraction of polar radio-metabolites in the brain limiting their use for in vivo PDE2A imaging. The latest radioligand out of this series, [18F]**TA5** ((IC50 PDE2A = 3.0 nM; IC50 PDE10A >1000 nM, Figure 1), exhibited the highest potency towards PDE2A and selectivity over PDE10A. However, autoradiographic studies of [18F]**TA5** showed a homogenous and non-displaceable binding in rat and pig brain cryosections, indicating an insufficient specificity of this radioligand [25]. In addition, [ 18F]**TA5** is degraded extremely fast in the mouse. Therefore, we decided to perform further structural modifications to develop fluorine-18 labeled PET tracers with improved metabolic stability for the molecular imaging of PDE2A.

**Figure 1.** Selected phosphodiesterase 2A (PDE2A) PET radioligands [10,19,20,23–25].

In our continuous effort to develop fluorine-18 labeled PET tracers dedicated to molecular imaging of PDE2A, we selected the benzoimidazotriazine (BIT) scaffold (Figure 1) as lead structure by replacing the pyrido ring of the **TA** compounds with a benzo ring [26,27]. As a first result, we very recently reported on the synthesis and inhibitory potency of nine new fluorinated derivatives based on this BIT scaffold [28]. In order to increase the metabolic stability of the corresponding PDE2A radioligands, the fluorine was introduced by substituting a fluoropyridine ring at the benzene moiety. Out of this series, derivative **BIT1** (Figure 1) was selected as the most suitable candidate for 18F-labeling and biological investigation (IC50 PDE2A = 3.33 nM; 16-fold selectivity over PDE10A) [28].

Herein, we reported on the development and the evaluation of [18F]**BIT1**, including the synthesis of the corresponding nitro precursor, the manual radiosynthesis, and the transfer to an automated synthesis module, and the subsequent in vitro and in vivo investigations.

#### **2. Results and Discussion**

#### *2.1. Precursor Synthesis and Radiochemistry*

#### 2.1.1. Synthesis of the Labeling Precursor **5**

The synthesis of the reference compound **BIT1** has been reported previously [28]. For an efficient radiosynthesis of [18F]**BIT1**, we selected the nitro precursor **5** because of its higher reactivity to nucleophilic aromatic substitutions in comparison to a bromo-substituted precursor [29,30]. As depicted in Scheme 1, **5** was prepared in four steps starting from the BIT key intermediate **1 [28]**. Firstly, Miyaura borylation of **1** with bis(pinacolato)diboron, using potassium acetate and [Pd(dppf)Cl2] as base and catalyst, afforded the pinacol boronic ester **2** in a satisfactory yield of 85% [31]. By palladium-catalyzed Suzuki coupling with 4-bromo-2-nitropyridine, compound **3** was obtained in 63% yield. This 4-bromo-2-nitropyridine was synthesized according to a slightly modified procedure, already described in the literature [32]. Thereafter, the bromination at 1-position of the imidazole ring using *N*-bromo-succinimide (NBS) afforded compound **4** in an excellent yield of 97%. Finally, the subsequent Suzuki coupling with 2-chloro-5-methoxy-phenyl boronic acid gave the nitro precursor **5** in 68% yield (NMR spectrums of precursor **5** see in the Supplementary Materials).

**Scheme 1.** Synthesis of precursor **5**. Reagents and conditions: (i) Bis(pinacolato)diboron [Pd(dppf)Cl2], KOAc, 2-methyltetrahydrofuran (2-MeTHF), 100 ◦C (6 h), 85 ◦C (12 h), 85%; (ii) 4-bromo-2-nitropyridine, K2CO3, Pd(PPh3)4, 1,2-dimethoxyethane, 63%; (iii) N-bromo-succinimide (NBS), acetonitrile (MeCN), room temperature (RT), 4–6 h, 97%; (iv) 2-chloro-5-methoxy-phenyl-1-yl-boronic acid, K2CO3, [Pd(PPh3)4], 1,4-dioxane/H2O (4:1), reflux, 1–2 d, 68%.

#### 2.1.2. Radiosynthesis and Characterization of [18F]**BIT1**

#### Manual Radiosynthesis

The new PDE2A radioligand [18F]**BIT1** was prepared by heteroaromatic nucleophilic substitution of **5** using a classical anhydrous K[18F]F-K2.2.2 carbonate complex. The radiolabeling process of [ 18F]**BIT1** was carried out under thermal heating applying a constant amount of 1.0 mg of precursor **5** by using different (i) polar aprotic solvents dimethylsulfoxide (DMSO), *N,N*-dimethylformamide (DMF), and acetonitrile (MeCN), (ii) temperatures, and (iii) reaction times.

As shown in Figure 2, after 5 min (min) reaction time, a radiochemical yield (RCY) of ~57% was obtained when DMF was used at 150 ◦C. According to radio-TLC, besides [18F]fluoride, three radioactive by-products were observed, which accounted for 10% of total radioactivity. When the temperature was reduced to 120 ◦C, an increase of the RCY up to 93% was observed at this early time point. However, the RCY decreased with increasing the reaction time, indicating a decomposition of [18F]**BIT1** under these conditions. The highest RCY in DMF (~94%) was achieved when the reaction temperature was further reduced to 100 ◦C. At this temperature, the product remained stable over 20 min reaction time. By contrast, almost no radiofluorination could be observed when MeCN was used (RCY ≤1%). With DMSO as a solvent, high RCYs (>85%) were obtained at 150 ◦C and 120 ◦C, which remained constant up to 20 min reaction time in contrast to the findings with DMF. A further decrease to 100 ◦C resulted in an increase of the RCY (>95%) after 10 min of reaction time. The precursor **5** was stable throughout the time of analysis, as proven by HPLC (data not shown). Based on these results, DMSO was selected for the production of [18F]**BIT1**.

Due to the similarity of the chromatographic behavior of the nitro precursor and the corresponding 18F-radiotracer, the use of a low amount of precursor is beneficial for subsequent isolation of the radiotracer via semi-preparative HPLC [33,34]. Accordingly, the further reduction of the amount of the nitro precursor **5** up to 0.5 mg was investigated, achieving an excellent RCY of ≥95% (DMSO, 100 ◦C, 5 min reaction time). The selection of a suitable column for semi-preparative HPLC was also somewhat crucial in order to have a separation of the radiotracer and its nitro precursor in a reasonable time [33]. According to our previous experiences [33], a slightly polar C18 phase turned out to be most appropriate for the isolation of [18F]**BIT1** using a mixture of MeCN/water and ammonium acetate as a buffer.

[ 18F]**BIT1** was successfully isolated under the aforementioned conditions and further purified by solid-phase extraction (SPE). After formulation in 0.9% saline containing 10% ethanol, [18F]**BIT1** was obtained with an RCY of 38% (decay corrected to the end of the bombardment, EOB), and molar activities in the range of 38 GBq/μmol (at the end of synthesis, EOS).

**Figure 2.** The radiochemical yield of [18F]**BIT1** depending on reaction time, temperature, and solvent. (Conditions: K[18F]F-K2.2.2, thermal heating, 1.0 mg of **5**). BIT, benzoimidazotriazine.

#### Automated Radiosynthesis of [18F]**BIT1**

By using the most appropriate conditions of the manual procedure, the automated radiosynthesis of [18F]**BIT1** (Scheme 2) was established using a TRACERlab FX2 N synthesis module (GE Healthcare, Waukesha, WI, USA). The detailed configuration is shown in theexperimental part. Briefly, the [18F]Fwas firstly trapped on an anion exchange cartridge and then eluted into the reactor using an aqueous potassium carbonate solution. The reaction took place with the [18F]F- /K2.2.2./K2CO3 system and the nitro precursor **5** (0.5 mg) in DMSO at 100 ◦C for 5 min. After the isolation of [18F]**BIT1** via semi-preparative RP-HPLC (Figure 3A), the product was purified via SPE on an RP cartridge and formulated in sterile isotonic saline containing 10% of EtOH. The total synthesis time of [18F]**BIT1** was approximately 75 min. Analytical radio- and UV-HPLC of the final product, spiked with the non-labeled reference compound **BIT1**, confirmed the identity of [18F]**BIT1** (Figure 3B). Finally, the radiotracer was obtained with a radiochemical purity of ≥99%, an RCY of 54 ± 2% (EOB, n = 3), and molar activities in the range of 155–175 GBq/μmol (EOS, n = 3).

**Scheme 2.** 18F-labeling of [18F]**BIT1.**

**Figure 3.** (**A**) Semi-preparative radio- and UV-HPLC chromatograms of [18F]**BIT1** (conditions: Reprosil-Pur C18-AQ, 250 × 10 mm, 46% MeCN/20 mM NH4OAcaq., 5.5 mL/min); (**B**) Analytical radio- and UV-HPLC chromatograms of the final product of [18F]**BIT1** spiked with the non-radioactive reference **BIT1** (conditions: Reprosil-Pur C18-AQ, 250 × 4.6 mm, gradient system with eluent mixture of MeCN/20 mM NH4OAcaq., 1.0 mL/min).

In vitro Stability and Lipophilicity of [18F]**BIT1**

The radioligand [18F]**BIT1** was stable (radiochemical purity <sup>≥</sup>99%) in phosphate-buffered saline (PBS), pig plasma, *n*-octanol, as well as in the saline formulation containing 10% ethanol at 40 ◦C for up to 1 h. The lipophilicity was determined by the shake-flask method in the *n*-octanol/PBS system. With a logD7.4 value of 1.81 <sup>±</sup> 0.05, [18F]**BIT1** falls within the range of radiotracers with optimal brain passive diffusion [35–37]. However, the calculated coefficient distribution value (ACD/Labs, Version 12.0, Advanced Chemistry Development, Inc.) displayed value of 4.03. The significant deviations between the calculation and experimental methods were observed in particular when the pattern of connectivity and non-bonded intramolecular interactions are not included in the applied database [38–41]. Moreover, the big discrepancy between the experimental and the calculated logD7.4 value was already observed for our PDE2 tracer [18F]**TA5** and has been previously discussed [25]. It is assumed that the experimentally determined higher hydrophilicity of [18F]**TA5** might be due to the solvation effect related to hydrogen bonding and ionization of the radioligand in the aqueous buffered system [25,35]. Whereas with the software-based determination, this effect may be underestimated [25]. Therefore, the calculated logD value is often higher than the experimentally determined value [25,35]. Since **TA5** and **BIT1** are rather structurally similar, we assume the apparent strong discrepancy of logD for [18F]**BIT1** might also be caused by the same reasons as those for [18F]**TA5**.

#### *2.2. In Vitro and In Vivo Characterization of [18F]BIT1*

#### 2.2.1. In Vitro Autoradiography of [18F]**BIT1**

To investigate the distribution of binding sites of [18F]**BIT1** in the brain, in vitro autoradiographic studies using cryosections of pig brain were performed. As depicted in Figure 4B, the distribution pattern of [18F]**BIT1** corresponds to the known spatial distribution of PDE2A with a high density of binding sites in the caudate nucleus (Cd), nucleus accumbens (Acb), cortex (Cx), and hippocampus (Hip) (Nissl staining of the corresponding slice is shown in Figure 4A). However, [18F]**BIT1** also binds to the non-PDE2A specific regions cerebellum (Cb) and thalamus (Th), indicating binding of [18F]**BIT1** to other targets.

**Figure 4.** In vitro autoradiography of [18F]**BIT1** in pig brain slices, (**A**) Nissl staining, (**B**) Total binding of 1.72 nM of [18F]**BIT1**, (**C**) Blocking studies (10 μM of **TA1**), (**D**) Self-blocking (10 μM of **BIT1**). Abbreviations: (Acb: nucleus accumbens; Cb: cerebellum; Cd: nucleus caudatus; Cx: cortex; Hip: hippocampus; Th: thalamus).

To verify these findings, blocking studies with 10 μM of **TA1** (a potent PDE2 ligand) [42] (Figure 4C) and **BIT1** (Figure 4D) were performed. The decrease of [18F]**BIT1** binding of ~50% and ~30% in PDE2A specific regions Cd and Acb, respectively, observed for both **TA1** and **BIT1** indicated in vitro specificity of the radiotracer. However, the simultaneous decrease of [18F]**BIT1** binding in the range of 20–30% in Cb by **TA1**, as well as **BIT1**, suggested further non-specific binding of the radiotracer. We hypothesized the high non-specific binding of [18F]**BIT1** could be related to the moderate selectivity of **BIT1** over PDE10A. Accordingly, these results limited the suitability of [18F]**BIT1** for in vitro molecular imaging of PDE2A.

#### 2.2.2. In Vivo Metabolism of [18F]**BIT1**

The in vivo metabolism of [18F]**BIT1** was investigated in plasma and brain homogenate obtained from mice at 30 min p.i.. Prior to the analysis with RP-HPLC, the samples were treated with a mixture of MeCN/H2O (4:1, v/v) to precipitate the proteins. The recovery of radioactivity in plasma and brain samples was 96% and 99%, respectively. Intact tracer accounted for 43% and 78% of total activity in plasma and brain, respectively, indicating higher metabolic stability of [18F]**BIT1** in comparison to our previous PDE2A radioligands [18F]**TA3**, [18F]**TA4**, and [18F]**TA5 [23]**. As shown in the chromatograms (Figure 5), the two radio-metabolites ([18F]**M1** and [18F]**M2**) found in the brain were also observed in the corresponding plasma sample, indicating their ability to penetrate the blood-brain barrier. For further clarification, samples obtained as previously described in our in vitro metabolism study with **BIT1** using mouse liver microsomes [28] were investigated similarly by HPLC, but with UV detection (Figure 5C). On the basis of the in vitro metabolites **M1** and **M2**, both elucidated by LC-MS in the mentioned study, we can conclude about the brain penetrating radio-metabolites [18F]**M1** and [18F]**M2** as products of *N*-oxidation or *C*-hydroxylation. Besides, some of the more polar radio-metabolites detected in plasma could only tentatively be assigned as formed by mono-oxygenation, di-oxygenation, reduction, and demethylation, but were not investigated further.

**Figure 5.** Representative analytical HPLC chromatogram: (**A**) plasma and (**B**) brain samples at 30 min p.i of [18F]**BIT1** in a CD-1 mouse; (**C**) sample from incubation of **BIT1** with mouse liver microsomes (MLM) in the presence of NADPH for 90 min at 37 ◦C (UV detection: 228 nm; axis of retention time was adjusted). Conditions: isocratic system, column: Reprosil-Pur C18-AQ, 250 × 4.6 mm; eluent: 42% MeCN in 20 mM NH4OAcaq., flow: 1.0 mL/min.

#### *2.3. In Vivo PET-MRI Studies of [18F]BIT1*

Dynamic PET-MRI studies were performed in female CD-1 mice after intravenous administration of [18F]**BIT1**. As reflected by the time-activity curves (TACs) shown in Figure 6, the radioactivity uptake in the whole brain, striatum, and cerebellum reached standardized uptake values (SUVs) of about 0.7 at 5 min p.i., indicating blood-brain barrier penetration. Since there is no difference in the TACs between the whole brain, striatum, and cerebellum, [18F]**BIT1** is assumed to possess low specific binding also in vivo.

**Figure 6.** Averaged time-activity curves of [18F]**BIT1** in CD-1 mice (n = 2) with standard uptake values (SUV) in the whole brain, striatum, and cerebellum.

To further investigate this hypothesis, blocking studies were performed by concomitant injection of **TA1** and [18F]**BIT1**. However, since no significant reduction of radioactivity uptake in the PDE2A-specific region striatum was detectable (data not shown), the high non-specific binding of [18F]**BIT1** already observed in vitro was confirmed.

Overall, the herein reported potential PDE2A radioligand, selected from a new class of 8-pyridinyl-BIT compounds, demonstrated sufficient blood-brain barrier (BBB) permeability. However, [ 18F]**BIT1** was found to be insufficient for in vivo imaging of PDE2A with PET. Further structural modifications are needed to obtain PDE2A selective radioligands for in vitro and in vivo research. Extensive structure-activity relationship (SAR) studies could lead to the improvement of the selectivity and specificity of compounds. In particular, modifications of the set substituents at 1- and 8-positions of the BIT scaffold might result in additive as well as nonadditive effects in compound potency [43].

#### **3. Materials and Methods**

#### *3.1. General Methods*

All chemicals were purchased from commercial sources and used without further purification. Solvents were dried before used if required. Air and moisture-sensitive reactions were carried out under argon atmosphere. Room temperature (RT) refers to 20–25 ◦C. Reaction mixtures were monitored by thin-layer chromatography (TLC) using pre-coated TLC-plates POLYGRAM® SIL G/UV254 (Macherey-Nagel GmbH & Co. KG, Düren, Germany). The spots were detected under UV light at λ 254 nm and 365 nm. For the purification of final products, flash chromatography was performed with silica gel 40–63 μm (Merck KGaA, Darmstadt, Germany). Radio-TLC was performed on Polygram SIL G/UV254 plates (Macherey-Nagel GmbH & Co. KG, Düren, Germany) pre-coated plates with a mixture of chloroform/methanol (10:1, v/v) as eluent. The radio-TLC plates were exposed to storage phosphor screens (BAS-IP MS 2025, FUJIFILM Co., Tokyo, Japan) and recorded using the AmershamTyphoon RGB Biomolecular Imager (GE Healthcare Life Sciences, Freiburg, Germany), and the images were quantified with ImageQuant TL8.1 software (GE Healthcare Life Sciences, Freiburg, Germany).

NMR spectra (1H and 13C) were recorded on Mercury 300/Mercury 400 (Varian, Palo Alto, CA, USA) or Fourier 300/Avance DRX 400 Bruker (Billerica, MA, USA) instruments. Residual solvent proton ( 1H) and carbon (13C) resonances were used as internal standards for 1H-NMR (CDCl3, δ<sup>H</sup> = 7.26; DMSO-*d6*, δ<sup>H</sup> = 2.50) and 13C-NMR (CDCl3, δ<sup>C</sup> = 77.16; DMSO-*d6*, δ<sup>C</sup> = 39.52). The chemical shifts (δ) were reported in ppm as follows: s, singlet; d, doublet; t, triplet; m, multiplet, and the coupling constants (J) are reported in Hz. Mass spectra were recorded on an ESQUIRE 3000 Plus (ESI, low resolution) and a 7 T APEX II (ESI, high resolution) (Bruker Daltonics, Bruker Corporation, Billerica, MA, USA).

For semi-preparative HPLC, the following conditions were used. Column: Reprosil-Pur C18-AQ, 250 × 10 mm, particle size: 10 μm; eluent: 46% MeCN/20 mM NH4OAcaq.; flow: 5.5 mL/min; ambient temperature; UV detection at 254 nm. The semi-preparative HPLC separations in the manual radiosynthesis were performed on a JASCO LC-2000 system, using a PU-2080-20 pump, an UV/VIS-2075 detector coupled with a radioactivity HPLC flow monitor (Gabi Star, raytest Isotopenmessgeräte GmbH, Straubenhardt, Germany) and a fraction collector (Advantec CHF-122SC, Dublin, CA, USA). Data recording was performed with the Galaxie chromatography software (Agilent Technologies, Santa Clara, USA).

Analytical chromatographic separations were performed on a JASCO LC-2000 system, incorporating a PU-2080*Plus* pump, AS-2055*Plus* auto-injector (100 μL sample loop), and a UV-2070 *Plus* detector coupled with a gamma radioactivity HPLC detector (Gabi Star, raytest Isotopenmessgeräte GmbH, Straubenhardt, Germany). Data analysis was performed with the Galaxie chromatography software (Agilent Technologies, Santa Clara, USA) using the chromatogram obtained at 254 nm. A Reprosil-Pur C18-AQ column (250 × 4.6 mm; 5 μm; Dr. Maisch HPLC GmbH, Ammerbuch-Entringen, Germany) with MeCN/20 mM NH4OAcaq. (pH 6.8) as eluent mixture and a flow of 1.0 mL/min was used (gradient: eluent A 10% MeCN/20 mM NH4OAcaq.; eluent B 90% MeCN/20 mM NH4OAcaq.; 0–10 min 100% A, 10–40 min up to 100% B, 40–45 min 100% B, 45–50 min up to 100% A, 50–60 min 100% A; isocratic system 42% MeCN/20 mM NH4OAcaq.; flow: 1.0 mL/min; ambient temperature).

The molar activities were determined on the basis of a calibration curve (0.1–6 μg of **BIT1**) performed under isocratic HPLC conditions (46% MeCN/20 mM NH4OAcaq.) using chromatograms obtained at 228 nm as the maximum of UV absorbance of compound **BIT1**.

No-carrier-added (n.c.a.) [18F]fluoride (t1/<sup>2</sup> = 109.8 min) was produced by irradiation of [18O]H2O (Hyox 18 enriched water, Rotem Industries Ltd., Arava, Israel) via [18O(p,n)18F] nuclear reaction by irradiation of on a Cyclone®18/9 (iba RadioPharma Solutions, Louvain-la-Neuve, Belgium).

Remote-controlled automated syntheses were performed using a TRACERlab FX2 N synthesizer (GE Healthcare, USA) equipped with a N810.3FT.18 pump (KNF Neuburger GmbH, Freiburg, Germany), a BlueShadow UV detector 10D (KNAUER GmbH, Berlin, Germany), NaI(Tl)- counter, and the TRACERlab FX Software.

#### *3.2. Precursor Synthesis and Radiochemistry*

The final compounds described in this manuscript meet the purity requirement (>95%) determined by UV-HPLC.

#### 3.2.1. Synthesis of Precursor

*3-Methyl-8-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)benzo[e]imidazo[5,1-c][1,2,4]triazine* (**2**)

A mixture of bromo derivative **1** [28] (1.32 g, 5 mmol), potassium acetate (1.1 g, 11.2 mmol), and bis(pinacolato)diboron (1.3 g, 5.11 mmol) in 2-MeTHF (20 mL) was degassed with argon for 10 min. After the addition of Pd(dppf)Cl2 (0.055 g, 0.075 mmol), the mixture was refluxed at 100 ◦C for 6 h and at 85 ◦C for 12 h. Upon completion of the reaction (monitored by TLC), CH2Cl2 (25 mL) was added, and the reaction mixture was stirred for 1 h. The solid was filtered off, and the filtrate was evaporated. The dark brown semi-solid residue (2.5 g) was dissolved in MTBE (60 mL) and filtered through a short silica gel column (H: 3 cm × D: 2 cm). Heptane (30 mL) was added to the eluate, followed by concentration to a volume of approximately 20 mL. The precipitated solid was filtered and dissolved

in MTBE (120 mL), and the solution obtained was filtered again through a short silica gel column (H: 2 cm × D: 2 cm). The yellow eluate was concentrated (→ ~15 mL) and treated with *n*-heptane (30 mL). The solid obtained was filtered and dried under reduced pressure to afford a yellow solid **2** (1.32 g, 85%). TLC [CHCl3/MeCN (10:1)]: Rf = 0.26 (strong tailing). 1H NMR (400 MHz, CDCl3) δ<sup>H</sup> = 8.57 (s, 1H), 8.41 (d, J = 8.0 Hz, 1H), 8.25 (s, 1H), 8.05 (dd, J = 8.0, 1.1 Hz, 1H), 2.89 (s, 3H), 1.40 (s, 13H).

*3-Methyl-8-(2-nitropyridin-4-5 yl)benzo[e]imidazo[5,1-c][1,2,4]triazine* (**3**)

A mixture of boronic ester **2** (470 mg, 1.5 mmol, 1 eq) and 4-bromo-2-nitropyridine (243 mg, 1.2 mmol) in 1,2-dimethoxyethane (6 mL) and an aqueous solution of K2CO3 (2 M, 1.75 mL, 3.5 mmol) was degassed with argon for 20 min. Pd(PPh3)4 (50 mg, 0.043 mmol) was added, and the reaction mixture was heated at 95 ◦C for 1 h and 88 ◦C for 14 h. Upon full conversion of starting material (monitored by TLC), the solvent was evaporated, and the residue was partitioned between water (30 mL) and CHCl3 (50 mL). After separation of the organic layer, the aqueous layer was extracted with CHCl3/MeCN (8:1, 6 × 30) and CHCl3 (5 × 30 mL). The combined organic layers were dried over Na2CO3, and the solvent was evaporated to leave a solid residue (370 mg). The residue was refluxed in MeCN (50 mL) for 5 min, and after cooling to 0–2 ◦C, the product was filtered and dried to give **3** as yellow solid (233 mg, 63%). TLC [CHCl3/MeOH/30% NH3 (10:1:0.1)]: Rf = 0.35. 1H NMR (300 MHz, DMSO-d6) δ<sup>H</sup> = 9.26 (s, 1H), 8.96 (dd, J = 2.0, 0.5 Hz, 1H), 8.85 (dt, J = 5.1, 0.8 Hz, 1H), 8.82–8.80 (m, 1H), 8.52 (dd, J = 8.7, 1.5 Hz, 1H), 8.42 (dd, J = 5.1, 1.7 Hz, 1H), 8.28 (dd, J = 8.5, 2.0 Hz, 1H), 2.78 (d, J = 1.7 Hz, 3H). 13C NMR (75 MHz, DMSO-d6) δ<sup>C</sup> = 157.67, 149.59, 149.28, 138.82, 138.14, 136.18, 135.67, 130.31, 127.22, 127.17, 126.02, 122.45, 115.55, 114.35, 12.18.

*1-Brom-3-methyl-8-(2-nitropyridin-4-yl)benzo[e]imidazo[5,1-c][1,2,4]triazine* (**4**)

*N*-Bromosuccinimide (NBS) (48 mg, 1.58 eq) was added to a suspension of compound **3** (52 mg, 1 eq) in 2.5 mL MeCN. The reaction mixture was protected from light and stirred at RT for 1–2 d. After the complete conversion of starting material (monitored by TLC), the mixture was partitioned between CHCl3 (15 mL) and water (15 mL). The organic layer was washed with aqueous NaHCO3 (5%, 15 mL) and Na2SO3 (5%, 15 mL). The aqueous layer was extracted again with CHCl3 (4x5 mL). The combined organic layers were dried (Na2SO4) and evaporated. The crude product was purified by column chromatography (CHCl3/MeOH, 20:1) to afford compound **4** as a yellow solid (64 mg, 97%). TLC [CHCl3/MeOH (20:1)]: Rf = 0.59. 1H NMR (400 MHz, CDCl3) δ<sup>H</sup> = 9.31 (d, *J* = 1.8 Hz, 1H), 8.84 (d, *J* = 5.0 Hz, 1H), 8.65 (d, *J* = 8.4 Hz, 1H), 8.59 (d, *J* = 1.6 Hz, 1H), 8.04–7.98 (m, 2H), 2.90 (s, 3H).

*1-(2-Chloro-5-methoxyphenyl)-3-methyl-8-(2-nitropyridin-4-yl)benzo[e]imidazo[5,1-c][1,2,4]triazine* (**5**) Brominated compound **4** (43 mg, 0.112 mmol, 1 eq), 2-chloro-5-methoxyphenyl boronic acid (27 mg, 0.145 mmol, 1.29 eq), and K2CO3 (44 mg, 0.318 mmol, 2.8 eq) were combined in a mixture of 1,4-dioxane and H2O (4:1). Under argon atmosphere, [Pd(PPh3)4] (13 mg, 10 mol%) was added. After refluxing at 101–105 ◦ C for 3 h, the reaction mixture was stirred at RT for 16 h. Afterward, the residue was partitioned between CHCl3 (20 mL) and water (20 mL). The organic layer was washed with water (20 mL) and brine (20 mL). The aqueous layer was extracted with CHCl3 (3 × 10 mL), and the combined organic layers were dried over Na2SO4, and the solvent was evaporated under reduced pressure. The crude product was purified by column chromatography using a gradient system (*n*-hexane/ethyl acetate from 1:1 to 1:3) to get the corresponding precursor **5** as a yellow solid (34 mg, 68%). TLC [*n*-hexane/ethyl acetate (1:2)]: Rf = 0.38. 1H NMR (400 MHz, CDCl3) δ<sup>H</sup> = 8.71 (dd, *J* = 5.0, 0.6 Hz, 1H), 8.60 (d, *J* = 8.4 Hz, 1H), 8.15 (dd, *J* = 1.6, 0.6 Hz, 1H), 7.95 (dd, *J* = 8.4, 1.9 Hz, 1H), 7.77 (dd, *J* = 5.0, 1.6 Hz, 1H), 7.67–7.64 (m, 1H), 7.58 (d, *J* = 1.9 Hz, 1H), 7.24–7.20 (m, 2H), 3.87 (s, 3H), 3.00 (s, 3H). 13C NMR (101 MHz, CDCl3) δ<sup>C</sup> = 159.24, 157.85, 150.52, 149.96, 140.32, 138.66, 137.53, 137.22, 136.47, 131.76, 131.27, 131.11, 126.50, 125.79, 125.61, 124.31, 118.59, 117.99, 115.77, 114.21, 56.06, 12.94. ESI (+): *m*/*z* = 447.10 (calcd. 447.10 for C22H16ClN6O3 <sup>+</sup> [M + H]+).

#### 3.2.2. In Vitro Metabolism of **BIT1**

**BIT1** was incubated with mouse liver microsomes in the presence of NADPH, including positive and negative controls, for 90 min at 37 ◦C, as previously described [28]. Obtained samples were investigated by analytical HPLC with an isocratic system (42% MeCN/20 mM NH4OAcaq; flow rate of 1.0 mL/min) and UV detection (228 nm).

#### 3.2.3. Radiochemistry

#### Manual Radiosynthesis

No carrier added [18F]fluoride in 1.5 mL water and was trapped on a Chromafix® 30 PS-HCO3 − cartridge (Macherey-Nagel GmbH & Co. KG, Düren, Germany). The activity was eluted with 300 μL of an aqueous solution of potassium carbonate solution (K2CO3, 1.8 mg, 13 μmol) into a 4 mL V-vial, and Kryptofix2.2.2. (K2.2.2) (11 mg, 29 μmol) in 1.0 mL MeCN was added. The aqueous [18F]fluoride was azeotropically dried under vacuum and nitrogen flow within 7–10 min using a single-mode microwave (75 W, at 50–60 ◦C, power cycling mode). Two aliquots of MeCN (2 × 1.0 mL) were added during the drying procedure, and the final complex was dissolved in 500 μL of solvent (DMSO, DMF, and MeCN) ready for labeling. Thereafter, 0.5–1.0 mg of precursor in 500 μL labeling solvent was added, and the 18F-labeling was performed at different temperatures (100 ◦C, 120 ◦C, and 150 ◦C). To analyze the reaction mixture and to determine radiochemical yields, samples were taken for radio-HPLC and radio-TLC at different time points (5, 10, 15, and 20 min).

After cooling to <30 ◦C, the reaction mixture was diluted with 1.0 mL water and 2.0 mL of MeCN/H2O (1:1) and directly applied to an isocratic semi-preparative RP-HPLC for isolation of [ 18F]**BIT1** (46% MeCN/20 mM NH4OAcaq., 5.5 mL/min, Reprosil-Pur C18-AQ column, 250 <sup>×</sup> 10 mm; 10 μm; Dr. Maisch HPLC GmbH, Ammerbuch-Entringen, Germany). The collected radiotracer fraction was diluted with 35 mL water to perform purification by sorption on a Sep-Pak® C18 light cartridge (Waters, Milford, MA, USA) and successive elution with 0.75 mL of ethanol.

#### Automated Radiosynthesis of [18F]**BIT1**

The automated radiosynthesis was performed in a TRACERlab FX2 N synthesis module (GE Healthcare, Waukesha, WI, USA). About 2–4.5 GBq of aqueous n.c.a [18F]fluoride was trapped on a Chromafix® 30 PS-HCO3 - cartridge (Figure 7, Macherey-Nagel GmbH & Co. KG, Düren, Germany, entry 1) and eluted with K2CO3 (1.8 mg, 13 μmol, entry 2) in 400 μL water. Kryptofix K2.2.2 (11 mg, 29 μmol, in 1.0 mL MeCN, entry 3) was directly added. Afterward, azeotropic drying was performed at 60 ◦C and 85 ◦C under a vacuum. Thereafter, 0.5 mg of **5** in 1.0 mL DMSO (entry 4) was transferred to the reactor, and subsequently, the reaction was performed at 100 ◦C for 5 min. After cooling the reactor to 30 ◦C, 1.0 mL of H2O and 2.0 mL of MeCN/H2O (1:1) (entry 5) were added and transferred into the injection vial (entry 6). The semi-preparative HPLC was performed using a Reprosil-Pur C-18 AQ column (250 × 10 mm; 10 μM) using 46% MeCN in aqueous 20 mM ammonium acetate at a flow rate of 5.5 mL/min (entry 7). The [18F]**BIT1** fraction was pooled in 35 mL H2O (entry 8) and then loaded on a preconditioned Sep-Pak® C18 light cartridge (Waters, Milford, MA, USA, entry 9), and washed with 2.0 mL of H2O (entry 10). Afterward, the trapped [18F]**BIT1** was eluted with 1.2 mL EtOH (entry 11) into the product vial (entry 12). The product was transferred out of the hot cell, and the solvent was reduced under a gentle argon stream at 70 ◦C to a final volume of 10–50 μL. Afterward, the radiotracer was diluted in isotonic saline to obtain a final product containing 10% of EtOH (v/v). The identity of the final product was confirmed using radioanalytical HPLC.

**Figure 7.** Scheme of the synthesis module TRACERlab FX2 N for the radiosynthesis of [18F]**BIT1**. (1) Chromafix® 30 PS-HCO3−, (2) K2CO3 (1.8 mg in 400 <sup>μ</sup>L water), (3) K2.2.2. (11 mg, 29 <sup>μ</sup>mol in 1.0 mL MeCN), (4) precursor (0.5 mg of **5** in 1.0 mL DMSO), (5) 1.0 mL H2O and 2.0 mL MeCN/H2O (1:1), (6) injection vial, (7) Reprosil-Pur C18-AQ column (46% MeCN/20mM NH4OAcaq., flow 5.5 mL/min), (8) 35 mL water, (9) Sep-Pak® C18 light, (10) 2.0 mL water, (11) 1.2 mL EtOH, (12) product vial.

The molar activity of [18F]**BIT1** was determined by analytical HPLC (isocratic mode: 46% MeCN/20 5mM NH4OAcaq., Reprosil-Pur C18-AQ column 250 × 4.6 mm, flow rate: 1.0 mL/min). The UV/Vis detection was carried out at 228 nm as the absorption maximum of **BIT1**.

#### Determination of Stability

In vitro stability of [18F]**BIT1** was investigated by incubation of the radioligand in phosphate-buffered saline (PBS, pH 7.4), *n*-octanol, and pig plasma at 40 ◦C for 60 min (~5 MBq of the radioligand was added to 500 μL of each medium). Samples were taken at 30 and 60 min and analyzed by radio-TLC and radio-HPLC.

#### Determination of log D

The lipophilicity of [18F]**BIT1** was determined by partitioning between *n*-octanol and phosphate-buffered saline (PBS, pH 7.4) at ambient temperature using the conventional shake-flask method. An aliquot of 10 μL of the formulated solution containing ~500 kBq of [18F]**BIT1** was added to a tube containing 6 mL of the *n*-octanol/PBS-mixture (1:1, v/v, four-fold determination). The tubes were shaken for 20 min using a mechanical shaker (HS250 basic, IKA Labortechnik GmbH & Co. KG, Staufen, Germany) followed by centrifugation (5000 rpm for 5 min) and separation of the phases. Aliquots of 1 mL of the organic and the aqueous phase were taken, and the activity was measured using an automated gamma counter (1480 WIZARD, Fa. Perkin Elmer, Waltham, MA, USA). The distribution coefficient (D) was calculated as [activity (cpm/mL) in *n*-octanol]/[(activity (cpm/mL) in PBS], specified as the decadic logarithm (logD).

#### *3.3. Animal Studies*

All experimental work, including animals, was conducted in accordance with the national legislation on the use of animals for research (Tierschutzgesetz (TierSchG), Tierschutz-Versuchstierverordnung (TierSchVersV)) and was approved by the responsible research ethics committee (TVV 18/18, DD24.1-5131/446/19, Landesdirektion Sachsen, 20th June 2018). Female CD-1 mice, 10–12 weeks, were obtained from the Medizinisch-Experimentelles Zentrum at Universität Leipzig. For the time of the experiments, the animals were kept in a dedicated climatic chamber with free access to water and food under a 12:12 h dark:light cycle at a constant temperature of 24 ◦C. Piglet brains were obtained from anesthetized and euthanized juvenile female German landrace pigs (Lehr- und Versuchsgut Oberholz, Universität Leipzig).

#### 3.3.1. In Vitro Autoradiography of [18F]**BIT1**

Pig brain cryosections (16 μm; Microm HM560 Cryostat, FischerScientific GmbH, Schwerte, Germany) were thawed, dried in a stream of cold air, and preincubated for 10 min with buffer (50 mM TRIS-HCl, pH 7.4, 120 mM NaCl, 5 mM KCl, 2 mM MgCl2) at room temperature. Afterward, brain sections were incubated with 1.72 nM of [18F]**BIT1** in the buffer for 60 min at room temperature. The sections were then washed twice with ice-cold 50 mM TRIS-HCl (pH 7.4), dipped in ice-cold deionized water, dried in a stream of cold air, and exposed for 60 min to an image plate. The analysis was performed using an image plate scanner (HD-CR 35; Duerr NDT GmbH, Bietigheim Bissingen, Germany). Non-specific binding was determined using 10 μM **TA1** and **BIT1** as blocking compounds.

#### 3.3.2. In Vivo Metabolism of [18F]**BIT1**

[ 18F]**BIT1** (~19 MBq) was injected in female CD-1 mice (10–12 weeks old) via the tail vein. Brain and blood samples were obtained at 30 min p.i., plasma separated by centrifugation (14,000 rpm, 1 min), and brain homogenized in ~1 mL isotonic saline on ice (10 strokes of a PTFE plunge at 1000 rpm) in borosilicate glass.

Protein precipitation of plasma and brain samples was performed in duplicate with ice-cold MeCN/H2O (8:2, v/v), which was added to the samples (sample/solvent, 1:4, v/v). The samples were vortexed for 1 min, followed by resting on ice for 10 min. Afterward, the samples were centrifuged for 5 min at 10,000 g. Supernatants were collected, and 100 μL ice-cold MeCN/H2O (8:2, v/v) was added to the pellets for the second extraction, applying the same treatment as before. The combined supernatants were concentrated at 70 ◦C under nitrogen stream until a remaining volume of 100 μL and subsequently quantified by analytical radio-HPLC with an isocratic system (42% MeCN/20 mM NH4OAcaq.; flow rate: 1.0 mL/min). The activity recovery was determined by measuring the radioactivity of aliquots taken from supernatants and the pellets using a gamma counter.

#### *3.4. PET-MRI Studies of [18F]BIT1*

PET/MRI scans were performed using a preclinical PET/MRI system (nanoScan, Mediso Medical Imaging Systems, Budapest, Hungary). Animals were anesthetized with a mixture of air/isoflurane (2.0%, 200 mL/min) (Anesthesia Unit U-410, agntho's, Lidingö, Sweden), and their body temperature maintained at 37 ◦C with a thermal bed system. [18F]**BIT1** (2 MBq, Am: 155 GBq/μmol, EOS) was injected via the tail vein, either without (baseline study, n = 2) or with **TA1** (2 mg/kg, DMSO/NaCl (1:25); blocking study, n = 2). A dynamic 60-min PET scan was started 20 sec before radioligand injection. The list-mode data were reconstructed into 33 frames (12 × 10 sec, 6 × 30 sec, 5 × 1 min, 10 × 5 min) with 3D-ordered subset expectation maximization, 4 iterations, and 6 subsets, using an energy window of 400 to 600 KeV and coincidence mode of 1–5. After the PET scan, an MRI scan was performed for anatomical orientation and attenuation correction on a 1T magnet using a T1-weighted gradient-echo sequence (TR = 15 ms, TE = 2.6 ms). The data analysis was done with PMOD (PMOD Technologies Ltd., Zurich, Switzerland, v. 3.6), and an atlas-based method was used to obtain SUV time-activity curves for the striatum, cerebellum, and the whole brain.

#### **4. Conclusions**

In our efforts to develop a specific radioligand for PET imaging of PDE2A in the brain on the basis of a benzoimidazotriazine scaffold, we successfully prepared the novel PDE2A radioligand [18F]**BIT1** with a high radiochemical yield, radiochemical purity, as well as molar activity. Our findings showed

that this class of compound demonstrated good brain penetration and sufficient in vivo metabolic stability. However, [18F]**BIT1** was not suitable for the molecular imaging of PDE2A in the brain because of a high non-specific binding. Further structural modifications are required to obtain more satisfactory PDE2A specific radioligands.

**Supplementary Materials:** The following are available online, Figures S1–S2: NMR spectra of precursor **5** ( 1H and 13C), Figure S3. The calibration curve of molar activity of reference **BIT1**.

**Author Contributions:** B.W., M.S., and P.B. designed the study; M.S. and R.R. designed and performed organic syntheses; B.W., R.R., and R.T. performed radiosynthesis; W.D.-C., B.W., R.T., F.-A.L., and R.R. performed in vivo metabolism studies; S.D.-S., W.D.-C., and P.B. performed in vitro autoradiographic studies; M.T., M.K., W.D.-C., and P.B. designed and performed PET-MR studies; R.R., B.W., R.T., M.T., M.K., W.D.-C., S.D.-S, F.-A.L., M.S., and P.B. wrote the paper. All authors read and approved the final manuscript.

**Acknowledgments:** This work was financially supported by the Deutsche Forschungsgemeinschaft (DFG), Project Number: SCHE 1825/3-1. We further thank the Research and Innovation in Science and Technology Project (RISET-Pro), the Ministry of Research, Technology, and Higher Education, Republic of Indonesia, World Bank Loan No. 8245-ID, for supporting the Ph.D. thesis of Rien Ritawidya, the staff of the Institute of Analytical Chemistry (University of Leipzig) for measuring NMR and LR/HR-MS spectra, Karsten Franke and Steffen Fischer for the production of [18F]fluoride, and Tina Spalholz for technical assistance.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Sample Availability:** Samples of the compounds are not available from the authors..

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **A Survey of Molecular Imaging of Opioid Receptors**

#### **Paul Cumming 1,2,\*, János Marton 3, Tuomas O. Lilius 4, Dag Erlend Olberg <sup>5</sup> and Axel Rominger 1,\***


Academic Editor: Peter Brust

Received: 21 October 2019; Accepted: 13 November 2019; Published: 19 November 2019

**Abstract:** The discovery of endogenous peptide ligands for morphine binding sites occurred in parallel with the identification of three subclasses of opioid receptor (OR), traditionally designated as μ, δ, and κ, along with the more recently defined opioid-receptor-like (ORL1) receptor. Early efforts in opioid receptor radiochemistry focused on the structure of the prototype agonist ligand, morphine, although *N*-[methyl-11C]morphine, -codeine and -heroin did not show significant binding *in vivo*. [ 11C]Diprenorphine ([11C]DPN), an orvinol type, non-selective OR antagonist ligand, was among the first successful PET tracers for molecular brain imaging, but has been largely supplanted in research studies by the μ-preferring agonist [11C]carfentanil ([11C]Caf). These two tracers have the property of being displaceable by endogenous opioid peptides in living brain, thus potentially serving in a competition-binding model. Indeed, many clinical PET studies with [11C]DPN or [11C]Caf affirm the release of endogenous opioids in response to painful stimuli. Numerous other PET studies implicate μ-OR signaling in aspects of human personality and vulnerability to drug dependence, but there have been very few clinical PET studies of μORs in neurological disorders. Tracers based on naltrindole, a non-peptide antagonist of the δ-preferring endogenous opioid enkephalin, have been used in PET studies of δORs, and [11C]GR103545 is validated for studies of κORs. Structures such as [11C]NOP-1A show selective binding at ORL-1 receptors in living brain. However, there is scant documentation of δ-, κ-, or ORL1 receptors in healthy human brain or in neurological and psychiatric disorders; here, clinical PET research must catch up with recent progress in radiopharmaceutical chemistry.

**Keywords:** opioid receptors; positron emission tomography; radiotracers; μOR-, δOR-, κOR- and ORL1-ligands; epilepsy; movement disorders; pain; drug dependence

#### **1. Introduction**

The analgesic and soporific properties of opium have been known since antiquity, perhaps first attested in the detached reveries of Homer's Lotophagi. The sinister side of opium dreams is depicted in Tennyson's version of that story, and more distinctly in the memoires of Thomas de Quincy, who may have had the distinction of establishing a genre of literature, the addiction diary. A key active constituent of the sap of *Papaver somniferum* was first isolated in 1804 by the apothecary Friedrich Wilhelm Sertürner, who named it morphium, later morphine (**1**). Chemists identified its elemental composition in the 19th century, and efforts to determine its structure were rewarded in 1925, when Gulland and Robinson [1] recommended a structure consistent with the characteristics of morphine and codeine and their degradation products. Subsequent investigations confirmed the correctness of the analytically deduced structure of morphine, culminating in its total synthesis, achieved in the 1950s by Gates and Tschudi [2,3]. The absolute stereochemistry of morphine's five chiral carbons (5, 6, 9, 13 and 14) was reported by Bentley and Cardwell [4] in 1955, and the first practically realizable morphine total synthesis with reasonable yields was reported by Rice in 1980 [5]. To this day, it is more economical to allow the poppy plant to do the main work of morphine (**1**) synthesis, although chemists have since produced so many structural variants that one might consider opioid pharmacology to be a discipline in its own right. There have been several reviews of opioid receptor imaging in the past decade [6–8], but we now present a comprehensive update on the the main classes of opioid receptor (OR) ligands used for positron emission tomography (PET), and review clinical findings with this technology. Relevant chemical structures of endogenous opioid peptides and representative small molecule opioid receptor ligands are depicted in Figure 1.

**Figure 1.** Chemical structures of endogenous opioid peptides and selected opioid receptor ligands.

The modern era of opioid pharmacology began with the identification of an opioid binding site in brain tissue in studies with tritiated naloxone [9]. Soon thereafter, opioid peptides were isolated from pig brain [10], which famously involved whisky as an emolument for the slaughterhouse workers. The pentapeptides Met5-enkephalin (**2**) and Leu5-enkephalin (**3**) both had morphine-like effects in inhibiting the electrically stimulated contraction of the *vas deferens*, with the latter compound being somewhat less potent. The enkephalins were most abundant in striatum and hypothalamus of rat, guinea pig and calf, and Met5-enkephalin (**2**) was generally 3–6 times more abundant than Leu5-enkephalin (**3**) [11,12]. An additional higher molecular weight opioid (β-endorphin) isolated and sequenced from camel pituitary extracts proved to be a 31 amino acid polypeptide possessing homology with Met5-enkephalin (**2**) [13]. A trypsin-sensitive opioid activity (dynorphin-A) isolated from pituitary is a 17 amino acid polypeptide possessing *N*-terminal homology with Leu5-enkephalin (**3**), with the shorter peptide dynorphin-B [1-13] having almost 1000-fold higher potency than Leu5-enkephalin (**3**) in the guinea pig ileum muscle preparation [14]. Soon after these discoveries, bovine DNA sequences were cloned for the β-endorphin precursor corticotropin-β-lipotropin [15], the Met5/Leu5-enkephalin (**2**, **3**) precursor preproenkephalin-A [16] (which proved to contain four copies of Met5-enkephalin (**2**) and one copy of Leu5-enkephalin (**3**), consistent with the ratio of their tissue concentrations), and the closely related preproenkephalin-B [17]. Other researchers cloned porcine preprodynorphin, the precursor for dynorphin A [1-17], dynorphin-A [1-8], dynorphin B [1-13], and other opioid peptides [18].

It was soon apparent that the endogenous opioid peptides bound to at least three distinct sites in the brain and peripheral tissues, known as μ-, δ-, and κORs. For a time, the orphaned σ receptors were thought to comprise another type of OR [20], due to the analgesic (and hallucinogenic) action of pentazocine at κORs in brain. However, binding of σOR ligands is not displacable by the opioid antagonist naloxone, nor do σORs bind opioid peptides with high affinity, such that the σ- receptor is now recognized as a pharmacological class in its own right. The μOR gene has at least 14 exons that can give rise to diverse splice variants, and at least three pharmacologically distinct subtypes are recognized: μ1, μ<sup>2</sup> and μ3. Displacement studies *in vitro* and *in vivo* with the μ-selective competitive antagonist cyprodime (**6**) and the μ1-specific competitor naloxonazine (**7**) showed that [11C]carfentanil ([11C]Caf, **8**) binds predominantly to the μ<sup>1</sup> subtype [21]. The μ<sup>3</sup> subtype is alkaloid sensitive and opioid peptide insensitive; it couples to nitric oxide generation, and, extraordinarily, its endogenous agonist in amygdala seems to be morphine (**1**) [22]. There is also an opioid receptor-like receptor 1 (ORL1), which is activated by the 17 amino acid polypeptide known as nociceptin/ orphanin FQ (N/OFQ) [23].

Opioid signaling has an important function in the modulation of pain processing at the spinal level. ORs synthetized in the dorsal root ganglion are transported to peripheral nerve endings and to the superficial layers of the spinal cord dorsal horn. In the dorsal horn, μORs are the most densely expressed subtype, followed by δ-, and κORs. Over 70% of the ORs there are located on the central terminals of small-diameter (mostly C and A-delta fibres) primary afferent neurons. A main mechanism of opioid analgesia lies in the activation of presynaptic μORs in the spinal cord, leading to decreased release of excitatory transmitters and nociceptive transmission. Unforunately, PET methods do not suffice for detecting ORs in the human spinal cord.

Figure 2 shows PET images of the distributions in human telencephalon of binding sites for the four main classes of ORs, μ, δ, κ, and ORL1. The pattern of μORs in Figure 2A encompasses the telencephalic pain pathway of limbic brain regions. Supraspinal μORs in the nucleus accumbens and amygdala have a role in the analgesic and reinforcing properties of opioids. The thalamus, especially the medial structures, relay nociceptive spinothalamic input from the spinal cord to higher structures. μORs also have a prominent distribution in the brainstem, with high density in several structures associated with analgesia, such as the periaqueductal gray, rostroventral medulla, the reticular formation, and locus coeruleus [24–26]. From these structures, efferent outflow descends to the spinal cord where it acts to inhibit nociceptive transmission in afferent fibres. μORs are also abundant in the hypothalamus, where they might affect hormonal regulation. Receptors in the medullary vagal complex, area postrema, and nucleus tractus solitarius, can mediate endocrine actions and nausea.

As seen in Figure 2B, the δOR has high expression in the cerebral cortex, nucleus accumbens, and the caudate putamen. This receptor is involved in analgesic activity at both spinal and supraspinal sites. Similarly as μORs, agonists of central δORs contribute to respiratory depression, whereas receptors in the gut mediate constipation, an important side effect of morphine. The δORs receptors localize presynaptically where they inhibit the release of excitatory neurotransmitters [27]. Despite these properties, δ-selective drugs have not yet found clinical application.

The κORs have wide expression in rat brain, with highest levels in the ventral tegmental area, substantia nigra, nucleus accumbens, caudate putamen, claustrum, endoperiform nucleus, various hypothalamic nuclei, and the amygdala [28]. A similar expression profile occurs in the human brain [29,30], as seen in Figure 2C. Activation of κORs does not produce respiratory depression, but typical adverse effects include sedation and dysphoria, limiting the clinical use of κOR targeting drugs [27]. Despite generally dysphoric effects in humans, the κOR agonist Salvinorin-A, which is obtained from the leaves of *Salvia divinorum*, finds a niche market in the drug subculture for those seeking to briefly experience a dissociative state.

**Figure 2.** Human brain opioid receptor PET images in coronal (upper row) and axial (lower row) planes. Modified and reproduced with permission from Peciña et al. [19]. From left to right, we see (**A**) the μOR agonist [11C]carfentanil, binding most abundantly in the caudate nucleus, anterior cingulate cortex, thalamus, and pituitary gland; (**B**) the δOR antagonist *N*1- -([11C]methyl)natrindole, which has diffuse binding throughout neocortex; (**C**) the κOR antagonist [11C]LY2795050, which has high binding in the insular cortex, lateral frontal cortex and amygdala; (**D**) the NOP antagonist [11C]NOP-1A, which binds abundantly throughout the brain. Binding sites of μ-, κ- and NOP-OR ligands are expressed as binding potential relative to the cerebellum (BPND), whereas binding of the δ-ligand (which has no non-binding reference region) is expressed as net influx (*Ki*,) in units of perfusion (mL cm−<sup>3</sup> min<sup>−</sup>1). The color scale in the lower right indicates (for **A**, **C**, and **D**) BPND ranging from 0 to 2, or (**B**) *Ki* ranging from 0–0.1 mL cm−<sup>3</sup> min−<sup>1</sup>

Figure 2D shows a widespread and abundant expression of NOP binding sites in human brain. Agonists of the NOP receptor, unlike μOR agonists, are devoid of reinforcing or motivational properties, but are implicated in homeostatic functions such as feeding and body weight, as well as anxiety, stress, and alcohol dependence [31].

#### **2. Radiotracers for the PET Imaging of ORs**

#### *2.1.* μ*OR Ligands and Non-Selective Ligands*

The era of OR PET imaging was preceded by a phase of studies *ex vivo* with tritiated ligands such as the antagonist [3H]diprenorphine ([3H]DPN), which accumulated in striatum, *locus coeruleus*, *substantia nigra pars compacta*, and *substantia gelatinosa* of the living rat [32]. A similar pattern was revealed with the antagoninst [3H]naloxone, which showed sodium-dependent saturable binding *ex vivo*, with a Bmax close to that seen *in vitro* [33]. In contrast, the agonist [3H]Foxy bound with low nM affinity at μORs *in vitro*, but failed to accumulate in brain of living rats, a property which was attributed to the presence of high sodium in the living organism. The presence of sodium in the biding medium enhanced antagonist binding *in vitro* but decreases agonist binding [34]. However, addition of

Na<sup>+</sup> to the incubation medium had little effect on the affinity of the morphiceptin analog μOR agonist Tyr-Pro-(*N*Me)Phe-D-Pro-NH2 *in vitro* ([3H]PL017) [35]. Unpredictable sensitivity of binding to the presence of sodium in the medium seems mainly to be a property of agonist ligands.

OR PET imaging began with the introduction of 3-*O*-acetyl-[18F]cyclofoxy (3-*O*-Ac-[18F]FcyF), (**10**). 3-*O*-Ac-[18F]FcyF (**10**, Figure 3) is an opioid antagonist radiotracer, which was prepared from 3-*O*-acetyl-6α-naltrexol triflate *via* direct nucleophilic substitution with tetraethylammonium [ 18F]fluoride in anhydrous acetonitrile at 80 ◦C for 15 min [36,37]. Based on displacement studies with CyF, binding of 3-*O*-Ac-[18F]FcyF (**10**) is likely to reveal the composite of μ- and κOR binding [38], despite the qualitatively μOR-like binding pattern reported in living baboon brain examined with 3-*O*-Ac-[18F]FcyF (**10**) [36], and the pattern of [3H]cyclofoxy retention in rat brain analysed *ex vivo* [39]. However, in a rat study, the increased [3H]cyclofoxy binding provoked by chronic treatment with morphine (**1**) could be attributed to upregulation of μOR sites [40]. Saturation binding PET studies with [18F]FcyF (**11**) in awake rat indicated a single binding site with apparent affinity of 2 nM and Bmax ranging from 15 pmol/g in white matter to 74 pmol/g in striatum; these results matched closely the corresponding results obtained *in vitro* [41].

**Figure 3.** Structures of μ-selective and non-selective opioid receptor radioligands.

PET studies with *N*-[methyl-11C]-labelled morphine, codeine, heroin and pethidine indicated distinct differences in uptake and kinetics in rhesus brain [42], all seemingly in relation to lipophilicity of the various drugs. A more detailed kinetics analysis of *N*-[methyl-11C]pethidine (**12**) in brain of rhesus monkey indicated a very low binding potential [43]. The methadone analogue [11C]*L*-α-acetylmethadol ([11C]LAAM, **13**) had moderate uptake in brain of mice, but its specific binding was not reported [44]. These observations are a case in point supporting the generalization that effective pharmaceuticals do not necessary make good ligands for molecular imaging of their targets.

The displaceability of OR ligands by competitors *in vivo* is a complex matter, and one increasingly relevant given the current opioid abuse crisis in some countries. Whereas antagonists such as naloxone are effective in rescuing addicts from death by overdose, naloxone also finds experimental use in

molecular imaging studies to confirm binding of PET tracers to ORs. Thus, the BPND of [11C]Caf (**8**) to μORs in human brain, which ranged from 1.0 in cerebellum to 2.7 in caudate nucleus, was nearly completely displaced throughout in brain by a 50-mg dose of naltrexone (NTX, **5**) [45], closely matching the dose used for resuce from opioid agonist overdose. In humans, intranasal naloxone administration caused a rapid displacement of [11C]Caf (**8**), in accordance with the rapid response seen in treatment for overdose [46].

Buprenorphine (BPN) [47–49] is a narcotic analgesic used since the 1970s in the low dose management of post-operative pain. Since 2002, BPN has approval in the United States at higher dose or in combination with naloxone (Suboxone®) for substitution therapy in the management of opiate addiction. BPN contains the same 6,14-ethenomorphinan skeleton as diprenorphine (DPN), and both compounds have an *N*17-cyclopropylmethyl substituent, although BPN contains a *tert*-butyl group in position-20 instead of methyl. Interestingly, BPN has a completely different pharmacological profile than DPN. Whereas DPN is a mixed antagonist, BPN is a partial μOR agonist and κOR antagonist, and displays some affinity for the NOP receptor [50] (Table 1).

**Table 1.** Binding profile of selected ligands at the human opioid receptors [51] .


EOP: Endogenous opioid peptide, EM: 4,5-Eopxy-morphinan, 4-AP: 4-Anilidopiperidine, orvinol: 6,14 ethenomorphinan, Bentley-compound, ArAP: Arylacetamidopiperazine, APPB: Aryl-phenylpyrrolidinylmethylphenoxy-benzamide, JDTic: *trans*-3,4-dimethyl-4-(3-hydroxyphenyl)- piperidine, **NND**: "non nitrogenous" diterpene, FDPTP: 2- -fluoro-4- ,5- -dihydrospiro[piperidine- 4,7- -thieno [2,3-c]pyran]- derivative, SPB: [[spiro[2.5]octan-8 yl]-methyl]piperidin-4-yl] benzimidazol-2-one, a: in the rat brain b: in guinea pig brain membranes.

Luthra et al. [73] synthesized *N*17-cyclopropyl[11C]methyl-buprenorphine starting from *N*17-*nor*-buprenorphine. Analogously to the *N*17-cyclopropyl[11C]methyl-diprenorphine synthesis [74] the corresponding precursor, *nor*-BPN, was reacted with cyclopropyl[11C]carbonyl chloride and the carbonyl functional group of the resulting intermediate was reduced with LiAlH4. Lever et al. [75] developed a metabolically stable radiotracer, 6-*O*-(methyl-11C)-BPN, at Johns Hopkins University in 1990 in a two-step synthesis from 3-*O*-TBDMS-6-*O*-desmethyl-BPN. The precursor was selectively alkylated in position-6 with [11C]iodomethane/NaH in DMF at 80 ◦C for two min. Following desilylation, [11C]BPN (**14**) was produced in 10% radiochemical yield with molar activity of 41 GBq/μmol. Subsequently, Luthra et al. [76], aiming to avoid the formation of 3-*O*-alkylated by-products, introduced the base-stable, acid labile trityl protecting group to protect the phenolic hydroxyl in position-3. Applying 3-*O*-trityl-6-*O*-desmethyl-BPN in a two-step, fully-automated

radiosynthesis (11C-methylation/deprotection), yielded [11C]BPN (**14**) in 15% radiochemical yield and with a molar activity of 13-22 GBq/μmol. In 2014, Schoultz et al. [63] reported a procedure for the radiosynthesis of 6-*O*-(2-[18F]fluoroethyl)-6-*O*-desmethyl-BPN ([18F]FE-BPN (**19**)) via 18F-fluoroalkylation of 3-*O*-trityl-6-*O*-desmethyl-BPN (TDBPN) precursor with [18F]fluoroethyl tosylate and subsequent trityl deprotection. The decay corrected formulated product yield was 26 % and the molar activity 50–300 GBq/μmol.

DPN, a semisynthetic thebaine/oripavine derivative with a methyl group amenable for labelling in position-20, belongs structurally to the ring-C bridged morphinans (6,14-ethenomorphinans, orvinols, Bentley-compounds) [48]. DPN is a nonselective OR antagonist with affinity in the nanomolar range (Table 1), 100 times more potent than nalorphine. Indeed, DPN is used in the veterinary medicine as an antidote/reversing agent/antagonist for remobilizing large African animals (rhinos/elephants, Revivon®), which had been immobilized with the astonishingly potent agonists etorphine or carfentanil.

The first attempt at labelling of DPN with carbon-11 in postition-20 was reported by Burns et al. [77], who used *N*-cyclopropylmethyl-dihydronororvinone as precursor. The reaction of the precursor bearing an acetyl group in position-7-alpha with [11C]methyllithium yielded 20-[11C]methyl-DPN ([11C]DPN (**15**)). Luthra et al. [74] developed cyclopropyl[11C]methyl-DPN by alkylating *N*17-nor-DPN with cyclopropane[11C]carbonyl chloride and then reducing the *N*-cyclopropyl[11C]carbonyl intermediate with LiAlH4 in THF. The radiochemical yield of the corresponding radioligands were low in both cases [74,77].

In 1987, Lever et al. [78] developed a [11C]DPN (**15**) synthesis by alkylating the precursor 3-*O*-TBDMS-6-*O*-desmethyl-DPN in position-6 with [11C]iodomethane in DMF containing sodium hydride at 80 ◦C for two min. After cleavage of the TBDMS protecting group, [11C]DPN (**15**) was obtained with 10 % radiochemical yield and 64 GBq/μmol molar activity. In 1994, Luthra and her associates at the Hammersmith Hospital developed a new precursor for the radiosynthesis of [11C]DPN (**15**) [76]. Selective alkylation of 3-*O*-trityl-6-*O*-desmethyl-DPN (TDDPN) with [11C]iodomethane in the presence of NaH/DMF (95 ◦C, five min). Upon deprotection with 2 M hydrochloric acid (95 ◦C, two min) the radiotracer **15** was obtained with a radiochemical yield of 13–19% and a molar activity of 16–24 GBq/μmol. Recently, Fairclough et al. [79] at the University of Manchester reported a modified synthetic method also starting from TDDPN, yielding [11]DPN (**15**) with ten times higher molar activity (240 GBq/μmol) [76] and a radiochemical yield of 32%. The non-selective OR partial mixed agonist/antagonist 6-*O*-(methyl-11C)-BPN ([11C]BPN, **14**) accumulated in striatum, thalamus and cingulate cortex in living baboon brain. Analysis of the dynamic PET data with a model assuming irreversible trapping gave a net blood-brain clearance (*Ki*) of about 0.064 mL cm−<sup>3</sup> min<sup>−</sup>1, which was halved by administration of naloxone, indicating substantial displaceability [80]. In a study in heroin addicts, the BPN occupancy at [11C]Caf (**8**) binding sites was estimated relative to the drug-free baseline. An oral dose of 2 mg BPN had an occupancy of about 50% throughout brain, whereas 16 mg had 85% global occupancy [81]. On the other hand, therapeutic methadone (18–90 mg/day) did not provoke any discernible occupancy at 6-*O*-(methyl-11C)-diprenorphine ([11C]DPN, **15**) binding sites, neither in human opioid addicts, nor in mice, a phenomenon attributed to high agonist potency of methadone, such that withdrawl effects are averted with a rather low occupancy [82]. In further preclinical studies from the same research group, binding of [11C]DPN (**15**) in mouse brain was unaltered by treatment with oxycodone (**4**) or morphine (**1**) (full agonists at μORs), but was reduced by approximately 90% by BPN (partial agonist at μORs and antagonist at the δ- and κORs).

A comparative OR PET study in humans compared the distributions of the μOR-selective agonist [ 11C]Caf (**8**) and the mixed antagonist [11C]DPN (**15**) [83]. Qualitatively, [11C]DPN (**15**) binding in the striatum, cingulate and frontal cortex exceeded that of [11C]Caf (**8**) (which had highest binding in the μOR-rich thalamus), consistent with labeling of additional non-μOR sites by [11C]DPN (**15**). An investigation of ORs in human cerebellum showed abundant binding of a μOR-specific ligand in the molecular layer, moderate binding of a κOR-selective ligand, but a near absence of δOR binding sites, which was consistent with the observations of [11C]DPN (**15**) binding *in vivo* [84]. The presence

of binding sites in cerebellum can be an obstacle to the valid use of reference tissue methods of PET quantitation.

As a longer-lived alternative to [11C]DPN, (**15**), Wester et al. developed 6-*O*-(2-[18F]fluoroethyl)-6-*O*-desmethyl-DPN ([18F]FE-DPN (**16**)) [85], which contains a 2-fluoroethoxy group in position-6 instead of an OCH3. [18F]FE-DPN (**16**) was synthesized from TDDPN, the so called *"Luthra-precursor"* [76], the same precursor as for [11C]DPN (**15**). For the synthesis of **16**, TDDPN was reacted with [18F]fluoroethyl tosylate ([18F]FE-Tos) in DMF in the presence of sodium hydride for five min at 100 ◦C. The trityl protecting group was removed with 2 N hydrochloric acid, yielding [18F]FE-DPN (**16**) with a radiochemical yield of 22 <sup>±</sup> 7% and the molar activity was 37 GBq/μmol [85]. In 2013, Schoultz et al. [86] reported an automated radiosynthesis of **16** from TDDPN with a decay-corrected radiochemical yield of 25 <sup>±</sup> 7%. [18F]FE-DPN (**16**) has similar uptake as [ 11C]DPN (**15**) in mouse brain, and obtained a BPND in human thalamus of about 2 relative to occipital cortex, versus only 0.3 in somatosensory cortex [87]. Women showed faster plasma metabolism [ 18F]FE-DPN (**16**) than men, which might contribute to apparent gender differences in binding [88].

The first instance of full compartmental analysis of an opioid PET ligand in living brain was for the case of [11C]Caf, as described below. This fentanyl analogue belongs to the 4-anilidopiperidine (4AP) class of OR ligands, which are potentμOR-selective agonists. Since 1960, numerous 4AP-type OR ligands were synthesized and their structure-activity-relationship at ORs were recently summarized [89,90]. Caf is structurally different from fentanyl in that it contains an additional carboxymethyl group in position-4 of the piperidine ring. Caf is a μOR-selective full agonist of extreme potency, being almost 10,000 times more potent than morphine (**1**) [58,91]. In 1985, [11C]Caf **(8)** radiotracer was applied in the first human PET study [59,92]. For the radiosynthesis of [11C]Caf, desmethyl-Caf sodium carboxylate was alkylated with [11C]iodomethane in DMF at 35 ◦C for five min [92]. This procedure gave molar activity at the end of synthesis of 122 GBq/μmol, which would correspond to mass dose of about 500 pg in a human PET study, which is too low to have any effect on particpants. According to a novel version of the radiosynthesis, desmethyl-Caf free acid serves as precursor in a reaction performed in dimethylsulfoxide with [11C]methyl triflate in the presence of tetrabutylammonium hydroxide [93]. This procedure gave a molar activity of 5 GBq/μmol, which would correspond to a mass dose of 100 μg, certainly intruding into the range causing some pharmacological effects. Risk of toxicity is a serious matter in PET imaging with potent agonists, and for society in general, given the weaponization [94] (figurative and literal) that is possible with Caf.

For the compartmental analysis of [11C]Caf (**8**), two models were fitted to dynamic time activity curves (TACs) measured by PET in human brain relative to a metabolite-corrected arterial input function [95]; this approach is the gold standard for PET quantitation. The input function obtained by HPLC analysis of plasma extracts showed that untransformed parent fractions declined to 50% at 25 min post injection. The authors estimated microparameters for the reversible transfer of the tracer across the blood-brain barrier (K1/k2), the reversible binding to a receptor compartment (k3/k4), and the reversible association to a non-specific compartment in brain (k5/k6). The mean binding potential (BP; k3/k4) was 1.8 in frontal cortex and 3.4 in thalamus at baseline, versus only 0.16 and 0.26 after treatment with naloxone (0.1 mg/kg). This study set a very high standard for quantitative PET analysis, although the molar activity of the tracer may not have been completely adequate. The amount of mass injected corresponded to about 5 μg Caf per scan, which could cause some analgesia in humans, although being less than 10% of the dose causing loss of consciousness. Nonetheless, this again raises the issue of safety in PET studies with the using of high-potency full agonist ligands, as noted above. Unless the the highest possible molar activity is obtained, pharmacologically significant occupancy can occur, bringing a risk of toxicity. Test-retest studies with [11C]Caf (**8**) showed admirable low variability (< 6%) and high intraclass correlation coefficients (ICC > 0.90) of the total distribution volume (VT) relative to the metabolite-corrected arterial input, and likewise BPND relative to a reference tissue [96].

Phenethyl-orvinol (PEO) [97] shares the same in ring-C bridged morphinan scaffold as DPN and BPN. PEO contains a 6,14-*etheno*-bridge, an *N*17-methyl substituent and a 2-phenethyl group in position-20. It is a full agonist with higher affinity at μOR (0.18 nM) and κORs (0.12 nM) than to δORs (5.1 nM). The radiosynthesis of 6-*O*-(methyl-11C)-phenethyl-orvinol ([11C]PEO, **18**) was reported by Marton *et al*. [62] in 2009. The *Luthra-type* trityl-protected precursor (3-*O*-trityl-6-*O*-desmethyl-phenethyl-orvinol, TDPEO) was alkylated in position-6 with [ 11C]iodomethane in the presence of 8-10 equiv. sodium hydride. The protecting group was removed with 1 M hydrochloric acid in ethanol, yielding [11C]PEO (**18**) with a radiochemical yield of 57 <sup>±</sup> 16% and a molar activity of 60 GBq/μmol.

In 2012, Marton and Henriksen [98] reported the preliminary results of the synthesis of 6-*O*-(2-[18F]fluoroethyl)-6-*O*-desmethyl-phenethyl-orvinol ([18F]FE-PEO, **17**) starting from 6-*O*-(2-tosyloxyethyl)-6-*O*-desmethyl-phenylethyl-orvinol (TE-TDPEO) *via* direct nucleophilic fluorination and subsequent deprotection. This procedure gave [18F]FE-PEO (**17**) in an isolated preparative yield of 35 ± 8% with a molar activity of 55–130 GBq/μmol. In 2013, a research group of the University of Cambridge [99] investigated [18F]FE-PEO (**17**) as a candidate OR PET-ligand, obtained by an automated cGMP-compliant method the [18F]FE-PEO at 28 <sup>±</sup> 15% yield and 52–224 GBq/μmol molar activity. In 2014, Schoultz et al. [63] reported the synthesis and biological evaluation of three structurally-related 6-*O*-(2-[18F]fluoroethyl)-6-*O*-desmethy-orvinols, i.e. [18F]FE-DPN (**16**), [ 18F]FE-BPN (**19**), and [18F]FE-PEO (**17**). The production of these 18F-fluoroethyl-orvinol radiotracers (**16**,**17**,**19**) was accomplished from 3-*O*-trityl-6-*O*-desmethyl-orvinol precursors (TDDPN, TDBPN, TDPEO) in a two-pot, three-step synthesis. [18F]FE-PEO (**17**) had a molar activity at end of synthesis of 50–250 GBq/μmol [99], corresponding to an injected mass <1 μg in human PET studies. The total distribution volume (VT) in rat brain ranged from 1 mL cm<sup>−</sup><sup>3</sup> in cerebellum to 8 mL cm−<sup>3</sup> in thalamus; displacement studies *in vitro* with the μOR-selective agonist DAMGO indicated high specificity in certain brain regions. [18F]FE-DPN (**16**) had a molar activity of 37 GBq/μmol [85].

#### *2.2. Delta Ligands*

*N*1- -Methylnaltrindole (MeNTI, Figure 4) is a highly selective δOR antagonist (Table 1). MeNTI was prepared from naltrexone (**5**) in a Fischer-indol synthesis with *N*-methyl-*N*-phenylhydrazine [64]. The radiosynthesis of [11C]MeNTI was reported by Lever et al. in 1995 [100]. In the first step, 3-*O*-benzyl-naltrindole was reacted with [11C]iodomethane in DMF in the presence of either sodium hydride or tetrabutylammonium hydroxide at 80 ◦C for two min. The next step was hydrogenolysis of the formed 3-*O*-benzyl-*N*1- -(methyl-11C)-naltrindole under heterogenous catalytic conditions (H2, 10% Pd/C, DMF/ethanol, 80 ◦C, four min), or alternatively catalytic transfer hydrogenation (HCOONH4, 10% Pd/C, MeOH). This gave [11C]MeNTI (**22**) with 6% radiochemical yield and a molar activity of 76 GBq/μmol.

**Figure 4.** Labeled δ-opioid receptor ligands.

Human PET studies with [11C]MeNTI (**22**) confirmed earlier demonstrations in living mice of δOR-selectivity *in vivo* [101]. The binding ratio relative to cerebellum ranged from 1.1 in hippocampus to 1.7 in striatum and insular cortex, regional values correlated rather precisely with known density of δORs *in vitro*, and showed 50% displacement after administration of 50 mg NTX (**5**). The tracer showed pseudo-irreversible binding characteristics over the course of 90 min, with net blood-brain

clearance (*Ki*) ranging from 0.04 in cerebellum to 0.11 mL cm−<sup>3</sup> min−<sup>1</sup> in putamen [102]. The *Ki* for [ 11C]MeNTI (**22**) in human brain declined by only about 20% after treatment with naloxone at a dose completely displacing μOR sites [45]. The authors of that study suggested that incomplete and variable δOR blockade might contribute to the success of NTX (**5**) as a treatment for alcoholism.

*N*1- -(2-[18F]fluoroethyl)naltrindole (**23**, [18F]FE-NTI, BU97001) was developed by Matthews et al. in 1999 [103]. The precursor, *N*1- -[2-(tosyloxyethyl)]-3-*O*-benzyl-naltrindole, was prepared in four consecutive transformations from naltrexone (**5**). In the first step, naltrexone (**5**) was reacted in a Fischer-indol synthesis with 2-(N*1*-phenylhydrazino)acetic acid ethyl ester. The resulting indolomorphinanyl-acetic ester was reacted with benzyl bromide to yield the 3-*O*-benzyl protected NTI derivative, which was reduced with LiAlH4 in THF-toluene to afford the corresponding indolomorphinanyl ethanol intermediate. This compound was reacted with tosyl chloride to provide the appropriate precursor with a tosyloxy leaving group. For the radiosynthesis of [ 18F]FE-NTI (**23**), the precursor was reacted in a direct nucleophilic reaction with potassium [ 18F]fluoride/K2CO3/Kryptofix[2.2.2] in DMF to yield *N*1- -(2-[18F]fluoroethyl)-3-*O*-benzyl-naltrindole. Final debenzylation by hydrogenolysis under heterogenous catalytical conditions (H2, Pd/C, *N*,*N*-dimethyl formamide) gave [18F]FE-NTI (**23**) with a radiochemical yield of 10% and molar activity of 31 GBq/μmol. [18F]FE-NTI (**23**) was an antagonist in mouse *vas deferens* with high selectivity over μ- and κOR sites, and its tritiated version bound to rat whole brain as a single site with KD of 0.42 nM and Bmax of 3 pmol g−<sup>1</sup> [104].

In 2007, Bourdier et al., [105] reported the radiosynthesis of a 2-[11C]methylpyrrolo[3,4-b]pyridine-5,7-dione derivative (*N*-substituted-[11C]quinolinimide) (**24**). The radiotracer containing a [11C]methyl-group on the pyridine ring was synthesized from a tributylstannyl precursor, with introduction of the [11C]methyl group by the Stille reaction using [11C]iodomethane in the presence of *tris*(dibenzylideneacetone)dipalladium, tri-*o*-tolylphosphine, K2CO3, and CuCl in DMF, heated at 90 ◦C for five min. The labelled compound (**24**) was synthesized with a radiochemical yield of 60 ± 10% and a molar activity of 30–56 GBq/μmol. The unlabelled version of the *N*-substituted quinolinimide had higher δOR-selectivity than MeNTI, but its 11C-derivative (**24**) failed to label ORs in mouse brain, due either to excessively rapid metabolism [105], or its only moderate affinity.

#### *2.3. Kappa Ligands*

GR89696 ((±)-methyl 4[(3,4-dichlorophenyl)acetyl]-3-(1-pyrrolidinylmethyl)- 1-piperazinecarboxylate [106–108] (Glaxo Group Research Ltd.) is a κOR-selective agonist with an arylacetamidopiperazine/diacylpiperazine structural core. GR103545, the biologically active (*R*)-(−)-enantiomer of GR89696, displays subnanomolar affinity and 1000-fold selectivity for human κOR (*Ki* = 0.02 nM), [65]. Ravert et al. [109] reported synthesis of both enantiomers of [11C]GR89696 (**26**, Figure 5) from the corresponding chiral normethylcarbamoyl precursor [108,110]. The radiosynthesis was accomplished by acylation of the secondary amine with [11C]methyl chloroformate in dichloromethane in the presence of trimethylamine, giving product with molar radioactivity of 69 GBq/μmol. Mouse brain distribution of the synthesized enantiomers, ((*R*)-(−)-[11C]GR103545 (**25**) and the (S)-(+)-enantiomer [11C]SGR) was determined *in vivo*, which showed the (*S*)-enantiomer to be inactive. The low radiochemical yield of the radiosynthesis (2–14%) [110,111] motivated the development of elaborate new radiochemical methods. In 2008, Schoultz et al. [112] developed a simple [ 11C]methyl triflate mediated methylation of carbamino adducts. Normethylcarbamoyl-GR103545 was converted to [11C]GR103545 (**25**) with [11C]CH3OTf under mild conditions in 64–91% radiochemical yield. Wilson et al. [113] developed a method for preparing [11C-carbonyl]-methylcarbamates directly from primary or secondary amines, applying either DBU or BEMP and cyclotron-produced [ 11C]CO2. [11C-carbonyl]-GR103545 (**25**) was synthesized with high radiochemical purity (>98%) and molar activity of 108–162 GBq/μmol. In 2011, Nabulsi et al. [114] reported an automated two-step, one-pot procedure for the synthesis of [11C]GR103545 (**25**) from normethylcarbamoyl-GR103545 *via* transcarboxylation using the zwitterionic carbamic complex DBU-CO2 and [11C]CH3OTf.

**Figure 5.** Selected κOR ligands.

In PET studies, the κOR-agonist ligand [11C]GR103545 (**25**) had a VT in baboon brain ranging from 3 mL cm−<sup>3</sup> in cerebellum to 10 mL cm−<sup>3</sup> in striatum and cingulate cortex [111]. Naloxone homogeneously displaced tracer throughout brain, but had no effect on VT in cerebellum, which would support use of that brain region as a reference tissue. The tracer had >100-fold selectivity for κ- over δand μORs *in vitro* [65]. Much as in baboons, PET studies in rhesus monkey showed VT ranging from 8 mL cm−<sup>3</sup> in cerebellum to 21 mL cm−<sup>3</sup> in striatum. Other monkey studies showed BPND ranging from 0.3 in amygdala to 2.2 in putamen [115]. This bolus plus infusion study with increasing mass dose in monkeys indicated an *in vivo* KD of 2 nM and Bmax of 1–6 pmol g−1. In 2014, Naganawa et al. [116] reported the first-in-human PET study with [11C]GR103545 (**25**). Test-retest variability of the quantitative endpoint VT was about 15%, and binding ranged from 8 mL cm−<sup>3</sup> in cerebellum to 41 mL cm−<sup>3</sup> in amygdala; Lassen plots with naltrexone blocking indicated a non-specific uptake (VND) of only 3.4 mL cm<sup>−</sup>3, thus emphasizing the absence of any non-binding reference region.

Based on the substituted-diacylpiperazine scaffold of GR103545, researchers at Yale University developed the new κOR agonist radiotracers [11C]EKAP (**27**) [68] and [11C]FEKAP (**28**) [67] with improved pharmacological and PET-imaging profile compared with the native compound. In the open-ring analogs of GR103545, the pyrrolidinyl-methyl group of the original molecule in position-3 was replaced by a diethylamino-methyl in EKAP and a ((ethyl)2-fluoroethyl)amino)methyl group in FEKAP. Imaging studies [11C]EKAP (**27**) in rhesus monkey showed rapid metabolism *in vivo* and fast, reversible binding kinetics in brain that was blockable with specific competitors. The BPND ranged from 0.8 in frontal cortex to 1.8 in globus pallidus.

Researchers at Eli Lilly, in cooperation with the Yale University, developed κOR antagonist radiotracers with the 3-pyridinyl-1-pyrrolidinylmethyl structural scaffold. Along these lines, in 2013, Zheng et al. [57] synthesized the selective κOR antagonist radiotracer [11C]LY2795050 (**30**) from an iodophenyl precursor in a two-step procedure. The precursor was converted by transition metal-mediated cyanation using H11CN and Pd2(dba)3/dppf to a [11C]nitrile intermediate. This latter was partially hydrolysed with NaOH/H2O2 in DMF at 80 ◦C for five min, giving a 12% radiochemical

yield with molar activity of 23.6 GBq/μmol. [11C]LY2459989 (**31**) was prepared in a two-step one-pot radiosynthesis. In the first step, an aryl-iodide-type precursor was transformed in a palladium catalyzed reaction (Pd2dba3/dppf) with H11CN to the corresponding [11C]nitrile, which was reacted with H2O2 under basic condition to afford **31** with 7.4% radiochemical yield and 23 GBq/μmol molar activity.

LY2459989 is the fluorine-containing analogue of LY2795050. Li et al. [117] synthesised the 18F-fluorine-labelled version of LY2459989 using two different methods. Using the nitro precursor, the radiochemical yield was too low, but applying the iodonium ylide precursor, [18F]LY2459989 (**30**) was prepared with 36% radiochemical yield and 1,175 GBq/μmol molar activity. While admirably high, this molar activity falls far short of the theoretical maximium for 18F-, which is 63,000 GBq/μmol. Where does all that non-radioactive fluoride come from?

As noted above, salvinorin A [118–120] is a naturally occurring non-alkaloid neo-clerodane diterpenoid, isolated from *Salvia divinorum*. Also as noted above, smokings Salvinorin-A can provoke a dissociative hallucinogenic experience distinct from that of the classical hallucinogens. It has a unique structure with seven chiral carbons and is potent and highly selective κOR agonist; salvinorin A does not display any significant activity at other OR subtypes. In 2008, Hooker et al. [121] synthesized the carbon-11 labelled version of salvinorin A (**33**), using salvinorin B as precursor for the radiosynthesis. The 2-alpha-hydroxyl group of the precursor was acylated with [11C]acetyl chloride in DMF in the presence of DMAP at 0 ◦C for 7–10 min, giving a radiochemical yield of 3.5–10% with molar activity 7.4–28 GBq/μmol. PET studies in baboon brain showed rapid uptake and washout of [11C]salvinorin A (**33**), matching the brief duration of the hallucinatory/dissociative experience reported by humans users. Rat studies showed that acute doses of salvinorin A caused dose-dependent occupancy at brain κORs labelled *in vivo* with [11C]GR103545 (**25**). Pretreatment with a high dose in the hours before PET examination caused persistent reductions in receptor availability, despite the brief plasma half-life of the drug, and despite the rather brief duration of the hallucinogenic experience. This suggests that κOR activation by agonists such as salvinorin A provokes a delayed and persistent receptor internalization [122].

In 2001, Thomas et al. [123] identified the first κOR-selective antagonist ligand, *trans*-3,4-dimethyl-4-(3-hydroxyphenyl)piperidine (JDTic, > 200-fold selective, Table 1), with non-opiate structure. MeJDTic is a derivative of JDTic that is ring methylated on the nitrogen of the tetrahydroisoquinoline. Poisnel et al. [69] prepared [11C]MeJDTic (**29**) from JDTic by methylation with [11C]methyl triflate in acetonitrile at room temperature for three min. The radiochemical yield was 78–98% and the molar activity 1.5–4.4 GBq/μmol. Recently, Schmitt et al. [124] synthesized *N*-[18F]fluoropropyl-JDTic ([18F]FP-JDTic) (**35**)) from JDTic with [18F]fluoropropyl-tosylate in DMSO in the presence of DIPEA and LiI at 150 ◦C for 30 min. *In vivo* studies in mouse showed accumulation of [ 18F]FP-JDTic (**35**) in peripheral organs rich in κORs. [11C]MeJDTic (**29**) entered mouse brain *in vivo*, albeit attaining a concentration of only 0.2–0.3% ID/g. Its binding was substantially reduced by treatment with the κOR agonist U50,488, but was unaffected by morphine (**1**) or naltrindole (NTI), thus attesting to its high selectivity for κOR sites [69]. The κOR-agonist [11C]LY2795050 (**30**) had an *in vitro* binding affinity of 1 nM, and 25 fold selectivity over μORs (Table 1). It readily entered monkey brain, and was substantially displacement by naloxone [57]. Displacement studies showed scant specific binding in monkey cerebellum, which supports its use as a reference region for quantitation. The tracer had a BPND as high as 1.0 in parts of the basal ganglia [125]. Dual tracer studies in monkey showed the LY2795050 displaced [11C]Caf (**8**) from μOR sites with an ED50 of 119 μg/kg, whereas the ED50 at κOR sites was 16 μg/kg, indicating 8-fold selectivity *in vivo* [125,126]. Corresponding human studies with kinetic modelling showed VT ranging from 2 mL cm−<sup>3</sup> in cerebellum to 4 mL cm−<sup>3</sup> in amygdala, and Lasson plots with partial NTX (**5**) blocking indicated a VND (non-specific binding) close to 1.6 mL cm<sup>−</sup>3, thus giving a BPND of 1.5 in amygdala versus only 0.2 in cerebellum [126]. The test-retest reliability in human brain was about 10% [127]. PET with [11C]LY2795050 (**30**) has revealed the dose-occupancy relationship in human brain for the experimental κOR antagonist LY2456302 (**34**), which is under development as a treatment of alcoholism [128].

[ 11C]LY2459989 (**31**) had sub-nM affinity at κOR sites *in vitro*, with 30-fold selectivity over μOR and 400-fold selectivity over δOR sites [66]. Preliminary PET studies in monkey showed rapid kinetics and substantial displaceability *in vivo*, with BPND ranging from 0.5 in thalamus to 2.2 in globus pallidus. A comparison of κOR ligands in rat showed that the agonist [11C]GR103545 (**25**) and the antagonist [ 11C]LY2459989 (**31**) had similar displacement by various κOR antagonists. However, and of great significance, the κOR agonists salvinorin A and U-50488, while displacing [11C]GR103545 (**25**) binding *in vivo*, did not alter [11C]LY2459989 (**31**) binding [129], which may indicate an allosteric binding mechanism. The novel κOR agonist tracer [11C]-EKAP (**27**) showed fast uptake kinetics and high specific binding in monkey brain, with VT ranging from 12 mL cm−<sup>3</sup> in cerebellum to mL cm−<sup>3</sup> in globus pallidus, corresponding to a BPND of 1.8, its binding was 95% displaced by pre-blocking with the antagonists naloxone or LY2795050 [68].

The highly selective and potent κOR-ligand U-50488 served as a scaffold for developing fluoro-alkylated PET ligands, but proved inappropriate due to 100-fold loss of affinity relative to the starting compound [130]. The novel fluorinated κ-ligand [18F]LY2459989 (**32**) had similar kinetic properties in monkey PET studies to those of [11C]LY2459989 (**31**) [117].

#### *2.4. Nociceptin and Opioid-like 1 Receptors (ORL1)*

Emerging evidence supports the use of agonists for the nociceptin/orphanin FQ peptide receptor (NOP) in the clinical management of pain and for substance abuse [131], thus presenting an attractive target for molecular imgaing A series of NOP ligands based on a 2- -fluoro-4- ,5- -dihydrospiro[piperidine-4,7- -thieno[2,3-c]pyran]-scaffold were screened in rats [71]. Uptake in monkey brain in a baseline condition contrasted with a blocking condition indicated specific binding of several of the [11C]-labelled compounds, of which [11C]NOP-1A ((2*S*)-2-[(2-fluorophenyl)methyl]- 3-(2-fluorospiro[4,5-dihydrothieno[2,3-c]pyran- 7,4- -piperidine-1- yl)-*N*-methyl-propanamide (**36**, Figure 6) was selected for further investigations. In the synthesis developed by Pike et al. [71], [11C]NOP-1A (**36**) was prepared from a primary-amide type precursor by methylation with [11C]iodomethane in DMSO basified with potassium hydroxide at 80 ◦C for 5 min. PET imaging experiments with **36** showed a VT in monkey brain ranging from 13 mL cm−<sup>3</sup> in cerebellum to 21 mL cm−<sup>3</sup> in amygdala. This fell globally to 8 mL cm−<sup>3</sup> after blocking with the antagonist SB-612111, indicating a BPND of 1–2 [132]. A somewhat lower VT range was detected in human brain [133], where the test-retest reliability was about 12% [134].

**Figure 6.** Chemical structures of selected ORL1 receptor ligands.

The NOP/ORL1 antagonist LY2940094 (**37**) exerted a dose-dependent reduction in immobility in the forced swim test, matching that provoked by imipramine, consistent with a potential antidepressant action [23]. Changes in [11C]NOP-1A (**36**) binding in brain of living rats revealed the ORL1 occupancy of orally administered LY2940094 (**37**) [135]. MK-0911 (1-(2-fluoroethyl)- 3-[(3*R*,4*R*)-3-(hydroxymethyl)-1-[[(8*S*)-spiro[2.5]octan-8-yl]methyl]piperidin-4-yl]benzimidazol-2- one) is a high affinity, selective NOP receptor antagonist developed by Merck Pharmaceuticals. The fluorine-18 labelled version [18F]MK-0911 (**39**) had a VT in human brain ranging from 5 mL cm−<sup>3</sup>

in cerebellum to 7 mL cm−<sup>3</sup> in temporal cortex, with excellent test-retest stability [72]. Displacement studies with antagonists revealed the presence of a small specific binding component in cerebellum, again raising a red flag about reference tissue quantitation. Studies with the nociceptine ligand [ 3H]PF-7191 (**38**) showed sub-nM binding displacement *in vitro* (*Ki* = 0.1 nM) and high selectivity over other OR types, as well as promising displaceable binding in rat brain measured *ex vivo* [136].

#### **3. Clinical Studies**

#### *3.1. Age and Gender*

The BPND of [11C]Caf (**8**) relative to occipital cortex was 20% lower in thalamus and amygdala of healthy, aged women compared with young women, but tended to increase with age in frontal cortex, whereas increases were more consistently seen in aged men [137]. This finding seems relevant to the age-dependent changes in sensitivity to μOR agonists, compounded by possible gender differences in hepatic tracer metabolism, noted above. Preliminary results with [11C]LY2795050 (**30**) PET did not indicate any change in κOR availability with age in humans [138]. Another κ-OR PET study with that ligand showed slightly higher (5–10%) VT in widespread brain regions of male subjects than that seen in age-matched women [139].

#### *3.2. Epilepsy*

A dual tracer PET study of patients with temporal lobe epilepsy showed increased binding of [ 11C]Caf (**8**) to μORs in the temporal neocortex and decreased binding in the amygdala ipsilateral to the epileptic focus [140]. However, binding of [11C]DPN (**15**) to μ- and other OR subtypes did not differ between affected and unaffected cerebral hemipsheres, emphasizing the importance of subtype selectivity in PET studies. Another multi-tracer PET study in temporal lobe epilepsy showed increased μOR binding ([11C]Caf (**8**)) in medial inferior temporal cortex and a more widespread increase in δOR binding ([11C]MeNTI (**22**)) in the affected temporal lobe [141]. Increased [11C]DPN (**15**) binding in temporal pole and fusiform gyrus of epilepsy patients declined with time since last seizure, indicating a transient response of the opioid system [142]. Applying a partial volume correction revealed small post-ictal increases in [11C]DPN (**15**) VT in the (sclerotic) hippocampus relative to the interictal state [143], possibly indicating reduced competition from endogenous opioids. Thus, there may be reduced opioid transmission in the post-ictal period.

A study of five patients with reading epilepsy (i.e. seizures provoked by reading text) revealed very circumscribed 10% reductions in [11C]DPN binding the temporal parietal cortex in the activation condition compared to non-reading baseline, whereas no such changes was seen in control subjects [144]. In the context of a competition model, the authors interpreted this to indicate task-dependent release of opioid peptides in the patients, but it is difficult to determine the causal relationship between this release and the seizures.

#### *3.3. Movement Disorders*

In an MPTP model of acquired parkinsonism, a substantial striatal dopamine depletion to FDOPA–PET was associated with a 20–30% reduction in the VT of [18F]FcyF, (**11**) in opioid-receptor rich areas, i.e., caudate, anterior putamen, thalamus, and amygdala relative to intact animals [145]. These animals had recovery of their motor function, suggesting that the μOR changes were an adaptive response to dopamine depletion. The same group reported that this effect was (paradoxically) bilateral in animals with unilateral dopamine lesions [146]. In patients with Parkinson's disease, the [11C]DPN (**15**) binding relative to occipital cortex was 20–30% reduced in striatum and thalamus only in those patients with iatrogenic DOPA-dyskinesia, but was unaffected in nondyskinetic Parkinson's disease patients [147]. This observation seems to merit further investigation, given the disabling effect of dyskinesias often encountered in the treatment of Parkinson's disease. On the other hand, there was no

difference in [11C]DPN (**15**) binding in symptomatic DYT1 primary torsion dystonia patients relative to controls [148].

Regional [11C]DPN (**15**) binding was unaffected in patients with restless legs syndrome (which, like, Parkinson's disease, is responsive to levodopa treatment). However, there was a negative correlation between VT and motor symptom severity, and a negative correlation between severity of pain and ligand binding in the medial pain system (medial thalamus, amygdala, caudate nucleus, anterior cingulate gyrus, insular and orbitofrontal cortex). The authors interpreted this result in relation to pain-induced release of endogenous peptides, rather than a primary aspect of restless legs syndrome [149].

Binding of the non-selective OR ligand [11C]DPN PET was reduced by 31% in the caudate nucleus and 26% in putamen of a small group of symptomatic Huntington's disease patients compared to age-matched healthy controls [150]. This effect was less pronounced than was the loss of dopamine transporters seen in the same patients, suggesting that the reduction in ORs may partially accommodate the nigrostriatal degeneration. Despite this finding, there has been no indication for opioid medications in the symptomatic treatment of HD.

#### *3.4. Pain*

A qualitative study of post pontine infarct central pain showed a reduction in [11C]DPN (**15**) uptake in the lateral cerebral cortex on the side contralateral to the main symptoms [151]. A more detailed study indicated [11C]DPN (**15**) binding reductions in contralateral thalamus, parietal, secondary somatosensory, insular and lateral prefrontal cortices; these reductions were similar irrespective of the site of the lesion causing the central pain syndrome [152]. Indeed, this network of brain regions is recognized as comprising a central pain pathway. Another [11C]DPN (**15**) study of central neuropathic pain showed 15–30% lower OR-binding within the medial pain system (cingulate, insula and thalamus), as well as the inferior parietal cortex of the lateral system (Brodman area 40). Patients with peripheral neuropathic pain had bilateral and symmetrical decreases in [11C]DPN (**15**) binding in contrast to the hemispheric changes seen in central pain patients [153]. As always, these binding reductions are ambiguous, perhaps due to reduced receptor expression, increased occupancy, or internalization.

Poor sleep quality in relation to topical application of 10% capsaicin cream (which directly activates cutaneous pain receptors) was associated with higher baseline [11C]Caf (**8**) BPND in the frontal lobe [154]. Thus, we suppose that baseline cortical binding may reflect a tradeoff between pain sensitivity and some cognitive or resilience function subserved by μ-ORs. Capsaicin-induced pain provoked a decrease in [11C]Caf (**8**) binding in the contralateral thalamus by as much as 50%, increasing as the subjective severity of the pain [155]. Heat pain reduced the [18F]FE-DPN (**16**) binding in limbic and paralimbic brain areas including the rostral ACC and insula [156]. Application of sustained painful stimulus of the jaw muscle with saline injection provoked bilateral reductions in [11C]Caf (**8**) binding in the ipsilateral amygdala (5%) and contralateral ventrolateral thalamus (7%) [157]. The same painful stimulus that provoked 5–10% decreases in [11C]Caf (**8**) binding in healthy young men tended to increase binding in women; this gender difference was most pronounced in the ventral striatum ipsilateral to the pain [158]. A small group of patients with trigeminal neuralgia had reduced [11C]Caf (**8**) binding in left nucleus accumbens, a brain region earlier implicated in pain modulation and response to reward or aversive stimuli [159]. It would be interesting to test if this phenomenon correlated with individual differences in affective state or trait neuroticism.

In another [11C]Caf (**8**) study, the reduced binding provoked soon after administration of a sustained pain of moderate intensity had normalized in the hours after cessation of the stimulus [160]. In general, painful stimuli do not desensitize with time, so the relationship between temporal dynamics of opioid signaling and pain perception must be complex. Indeed, pain researchers and clinicians alike are familiar with the phenomenon of allodynia, which is a decrease in pain threshold, or the conversion of previously non-painful stimuli to pain. In a sciatic nerve stimulation model, pain conditioning some hours after stimulation was associated with increased C-fibre response and reduced C-fibre threshold,

as well as supraspinal changes marked by increased binding of [11C]PEO (**18**) in ipsilateral and bilateral structures of the rat brain [161]. Thus, allodynia may indicate inactivation of pain-mediated opioid release in brain, resulting in greater OR availability. On the other hand, an [18F]FE-DPN (**16**) PET study in athletes contrasting receptor availability at rest with the condition immediately after running a half marathon showed reduced binding in various paralimbic and prefontal cortical structures, to an extent correlation with post-running euphoria ("runner's high") [162]. Similarly, a [11C]Caf (**8**) study in recreationally active men showed a relationship between post-exercising euphoria with decreased μOR binding in widespread cortical areas after high intensity exercise, although effects were less clear after moderate intensity exercise [163].

Prolonged electrical stimulation of the motor cortex for relief of neuropathic pain caused reductions in [11C]DPN (**15**) binding in part of the cingulate cortex, prefrontal cortex, the periaqueductal gray prefrontal cortex, and cerebellum [164]. Some of these changes correlated with the extent of pain relief. In a case study, a single session of motor cortex stimulation improved the cold pain threshold, while decreasing [11C]Caf (**8**) binding in pain-network brain regions [165]. Other studies showed that low cerebral binding of [11C]DPN (**15**) predicted for poor response to motor cortex stimulation for the treatment of neuropathic pain [166]. Stimulation of the central grey for treatment of phantom limb pain provoked a focal decrease in midbrain [11C]DPN (**15**) binding, indicating endogenous opioid release [167].

Visceral pain applied by gastric inflammation was without effect on cerebral [11C]Caf (**8**) binding in healthy volunteers [168], this standing in contrast to findings with somatic pain as described above. On the other hand, vestibular stimulation provoked decreased [18F]FE-DPN (**16**) binding in parts of the right dominant cortical vestibular network [169].

A [11C]Caf (**8**) study in which painful heat was applied after administration of supposedly analgesic cream indicated a placebo-mediated reduction in receptor availability [170]. The same group later showed that placebo transcranial direct current stimulation (tDCS) reduced [11C]Caf (**8**) BPND in the periaqueductal gray matter (PAG), precuneus, and thalamus, indicating endogenous opioid release [171]. This placebo effect apparently increased upon administration of verum tDCS. In another study, acupuncture administered according to an authentic analgesic procedure had only slight effects on the binding of [11C]DPN (**15**) in human brain [172]. However, a study with [11C]Caf (**8**) showed acupuncture therapy to provoke short-term and persistent 10–30% increases in μOR binding in pain-related brain regions; importantly the verum acupuncture condition was contrasted with a sham acupuncture condition in that study [173]. Thus, while acupuncture analgesia may be "in one's head", there seems to be a real component mediated by increased opioid transmission. Transcutaneous electrical acupoint stimulation (TEAS) is an analogue of the electrical acupuncture technique. Administration of TEAS at 2 Hz to anesthetized monkey provoked reductions in [11C]Caf (**8**) binding in striatum, amygdala and ACC, whereas 100 Hz stimulation had no effect relative to baseline PET recordings [174].

Reminiscent of the findings in the study of pain-induced sleep disturbance noted above, a cross sectional study of sensory processing in healthy volunteers showed lower baseline binding of [18F]FE-DPN (**16**) in regions such as insular cortex and orbitofrontal cortex of those with greater sensitivity to cold pain. In addition, there were negative correlations between regional binding and sensory thresholds for non-painful stimuli [175]. Similarly, the individual striatal binding of [ 11C]Caf (**8**) BPND predicted cold pressor pain threshold, but not cold pressor pain tolerance or tactile sensitivity [176]. A longitudinal [18F]FE-DPN (**16**) PET study in neuropathic pain model rats showed lower μ+κ OR availability in the insula, caudate putamen, and motor cortex at three months after the injury [177]. These reductions occurred in association with anhedonia (disinterest in sucrose solution). Overall, these studies suggest that individual differences in OR signaling may mediate vulnerability to environmental stressors, a topic to be elaborated in Section 3.5 below.

Binding of [11C]DPN (**15**) was reduced in pineal gland (but not in the brain *per se*) of patients who had been experiencing cluster headache attacks [178], said to be one of the most painful experiences. The authors suggested that inputs from trigeminal nerve to the pineal gland might mediate this change. A group of seven spontaneous migrainers showed ictal reductions in [11C]Caf (**8**) binding in the medial prefrontal cortex, which correlated with the baseline or intra-ictal binding [179]. The [11C]DPN (**15**) BPND in brain did not differ between migrainers and healthy controls, nor was there any effect of placebo treatment in either group [180].

#### *3.5. Personality, Drug Dependence, and Psychiatric Disorders*

Scores in the personality trait of harm avoidance in a group of 23 healthy males correlated positively with binding of [18F]FE-DPN (**16**) in the bilateral ventral striatum, suggesting a link with predisposition to substance abuse [181]. It might follow that drug abuse is a kind of self-medication for those with pronounced harm avoidance trait. A comparison of [11C]Caf (**8**) uptake in healthy individuals showed that high scores in the harm avoidance trait were associated with high μOR availability in frontal and insular cortex [182], again linking the hard avoidance trait with lower tonic opioid transmission. Score in a scale of behavioral activation, which conceptually guides approach behavior, and notably in a scale designated "fun-seeking", correlated positively with [11C]Caf (**8**) in widespread brain regions [183]. A [11C]Caf (**8**) study in 49 healthy volunteers showed an *inverse* relationship between μOR availability in various brain regions and individual scores in the avoidance dimension of interpersonal attachment [184]. Considering the harm avoidance findings, baseline μOR availability may mediate a trade-off between harm avoidance and avoidant behavior in interpersonal relationships, in a psychological analogue of pain or cold sensitivity.

In a large group of healthy women, [11C]Caf (**8**) binding had a negative correlation with BOLD signal responses in amygdala, hippocampus, thalamus, and hypothalamus to viewing emotionally arousing scenes [185]. Non-sexual, albeit pleasurable social touch from a partner provoked widespread increases in [11C]Caf (**8**) binding, suggesting reduced opioid signaling [186], whereas social laughter provoked by viewing comedic film clips decreased [11C]Caf (**8**) binding in thalamus, caudate nucleus, and anterior insula. Furthermore, baseline μOR availability in some regions was associated with the rate of social laughter [187]. These results are difficult to reconcile, since pleasurable social experiences can seemingly have opposite effects on μOR availability. Contrasting the [11C]Caf (**8**) binding in euthymic and unhappy states (provoked by autobiographical reflection) in young women showed higher μOR availability in the rostral anterior cingulate, ventral pallidum, amygdala, and inferior temporal cortex in the unhappy state [188]. This kind of sad reflection provoked greater increases in [ 11C]Caf (**8**) binding in widespread brain regions of women with major depression [189], suggesting an exaggerated opioid response in relation to mood disorder, as distinct from ordinary sadness. A pilot PET study with the κOR-ligand [11C]GR103545 (**25**) did not reveal any binding differences between healthy control and patients suffering from major depression [190]. However, a [11C]EKAP (**27**) κOR study in healthy volunteers showed an inverse correlation between social status and [11C]salvinorin A (**33**) binding in widespread brain areas, with a special association occurring in brain regions mediating reward or aversion [191]. Given the association between social stress and depression, one might have expected covariance κORs in the two studies.

A recent [11C]Caf (**8**) PET study of 19 schizophrenia patients and 20 controls showed a 10% lower (Cohen's d = 0.7) μOR -availability in striatium of the patient group. While such a decrease can hardly be pathogonomic of disease, the authors also reported considerably higher inter-regional covariance of the [11C]Caf (**8**) binding in the patients, which might indicate an aberent spatial pattern of opioid signalling in schizophrenia [192]. There have been no OR PET studies in bipolar disorder.

A [11C]Caf (**8**) study showed that circulating levels of the anti-nociceptive cytokine IL-1ra (which correlated with neuroticism scores) predicted for the pain response to a standard stimulus (saline infusion to the masseter muscle), and likewise the reduction in μOR availability in the basal ganglia during the painful stimulus [193]. In a group of female patients suffering from fibromyalgia, [11C]Caf (**8**) binding correlated with pain-evoked BOLD signal changes in several brain regions, including dorsolateral prefrontal cortex and nucleus accumbens [194]. Overall, these studies suggest some linking between opioid transmission, mood, and inflammatory markers, which returns to the the notioin that OR signaling may mediate personality traits and vulnerability to stresses of various sorts.

[ 11C]Caf (**8**) PET showed persistently increased μOR binding in frontal and cingulate cortex of acutely detoxified cocaine addicts, which correlated with the extent of craving [195]. Elevated [11C]Caf (**8**) binding in frontal and temporal cortical regions was a significant predictor of time to relapse to cocaine use among abstinent addicts [196]. Binge cocaine users showed a significant association between [11C]GR103545 (**25**) binding to κORs with the amount of drug consumed. Furthermore, a three-day cocaine binge reduced binding by about 15% [197]. The cerebral binding (VT) of the ORL1 ligand [11C]NOP-1A (**36**) was globally elevated 10% in detoxified cocaine users [198].

One [11C]Caf (**8**) PET study showed persistently increased μOR binding in striatum of detoxified alcoholics, which furthermore correlated with the extent of craving [199]. Abstinent alcoholics showed significantly higher [11C]Caf (**8**) binding compared to controls, but a blunting of the response to amphetamine (which indirectly displaces μOR binding), resembling that seen by the same research group in compulsive gamblers [200]. However, others saw only a small increase in [11C]DPN (**15**) VT in brain of acutely withdrawn alcoholics, although there was a correlation between individual PET results and craving scores at the time of scanning [201]; the combined (μ+κ) PET signal in that study makes difficult a comparison with [11C]Caf (**8**) studies*. Post mortem* autoradiographic examination of [ 3H]DAMGO binding in brain of a large group of alcoholics showed substantial reductions in μOR binding sites, whereas low [11C]Caf (**8**) BPND in ventral striatum of acutely detoxified patients predicted high risk of relapse and poor response to naloxone in interaction with the μOR rs1799971 allele [202]. The disagreement between μOR findings *in vivo* and *post mortem* could indicate low tonic occupancy in alcohol dependent patients, since competition effects would disappear in autoradiographic studies.

In a [11C]MeNTI (**22**) PET study, there was globally 10–20% higher δOR binding in brain of a large group of alcohol-dependent subjects; which reached significance upon correcting for age, gender, and smoking status; there was an inverse relationship between binding in some regions and intensity of craving [203]. The VT of the κOR-selective ligand [11C]LY2795050 (**30**) was significantly lower in amygdala and pallidum of alcohol-dependent subjects [138]. This stands in contrast to the usual finding of increased μOR binding and the single report of elevated δOR binding.

The naloxone challenge paradigm has a long history in investigations of the regulation of the neuroendocrine axis, but it has been uncertain if naloxone-induced increases in ACTH and cortisol secretion bear any relation to central OR pathways. In a [11C]Caf (**8**) PET study of healthy volunteers there were negative relationships between cortisol (but not ACTH) response to naloxone and ligand BPND in ventral striatum, putamen and caudate [204]. The inverse relationship between naloxone-induced cortisol secretion and [11C]Caf (**8**) BPND in various brain regions of healthy volunteers was absent in alcohol dependent subjects [205]. This suggests that central ORs exert a top-down regulation of the neuroendocrine axis, which might contribute to individual differences in stress response, and that the normal regulation of this axis is disprupted in alcohol dependence.

There was only a slight difference in [11C]Caf (**8**) binding between non-smoking carriers of the μOR rs1799971 allelic variants, but this allelic difference was greater among smokers. Furthermore, the contrast in PET results between active and denicotinized cigarette conditions revealed a positive relationship between reward and altered μOR availability in the smokers [206]. An apparent re-analysis of the same data showed widespread reductions in [11C]Caf (**8**) binding after smoking a nicotine-containing cigarette; this effect was moderated by the rs1799971 polymorphism, where carriers of the A allele showed greater response to active cigarette smoking, and higher baseline μOR binding [207]. The authors conceded that non-nicotinergic factors, i.e. conditioning, could be contributing to aspects of smoking related opioid transmission [208].

Smoking subjects with higher dependence, craving, and cigarette consumption rates showed lower baseline [11C]Caf (**8**) BPND in limbic brain regions. There was bluniting of this association during NTX (**5**) treatment [209], but there was very low residual specific binding in the NTX condition, which must have compromised the sensitivity of the method. Another [11C]Caf (**8**) PET study showed no difference in BPND between placebo and active nicotine cigarette conditions, and no difference between smokers and nonsmokers. However, there was a negative correlation in the smokers between BPND in bilateral superior temporal cortex with scores in an index of nicotine dependence [210].

As noted above, challenge with amphetamine can indirectly provoke increased opioid peptide release. However, in a placebo-controlled, double-blinded and randomized [11C]Caf (**8**) PET study, amphetamine challenge (0.3 mg/kg) did not alter μOR availability in healthy male volunteers [211]. This stands in contrast to another study, wherein a high dose of amphetamine (0.5 mg/kg) provoked reductions in [11C]Caf (**8**) binding in widespread brain regions, i.e. frontal cortex, putamen, caudate, thalamus, anterior cingulate, and insula, whereas a sub-pharmacological dose was without such an effect [212,213]. Preclinical studies point to the importance of receptor internalization on the vulnerability of OR-receptor binding to challenge with amphetamine [214]. Notwithstanding this caveat, amphetamine induced reductions in [11C]Caf (**8**) binding were blunted in compulsive gamblers compared to that in a healthy control group controlled for smoking and drinking [215]. There was a general correlation between dopamine synthesis capacity to FDOPA PET and [11C]Caf (**8**) binding in putamen of healthy controls, and likewise in pathological gamblers, consistent with a tight relationship between dopamine and opioid systems in relation to compulsive behaviors [216].

Women with bulimia nervosa had reduced [11C]Caf (**8**) binding in the left insula, to an extent correlating with their duration of fasting [217]. Obese patients (BMI 40) had globally 20% lower [ 11C]Caf (**8**) BPND compared to lean volunteers; contrary to some reports, the same obese subjects had normal dopamine D2 receptor levels in striatum [218]. There were similar reductions in [11C]Caf (**8**) binding in morbidly obese subjects and patients with binge eating disorder [219]. Weight loss after bariatric surgery for the treatment of obesity resulted in a global 25% increase of μOR binding [220]. A dual tracer study with [11C]Caf (**8**) and the dopamine receptor ligand [11C]raclopride showed a high correlation in the striatum of lean subjects, whereas this correlation was considerably weaker in the ventral (limbic) striatum of the morbidly obese, suggesting an uncoupling of opioid/dopamine interactions in that condition [221]. This finding might predict analogous results in gambling and substance abuse disorders, which likewise may involve dysregulation opioid/dopamine interactions.

Feeding, regardless of the hedonic experience (palatable versus unpalatable meal), provoked widespread decreases in [11C]Caf (**8**) binding in non-obese healthy subjects, suggesting that OR transmission mediates some aspect of the rewarding properties of food [222]. Also in non-obese subjects, [11C]Caf (**8**) BPND in amygdala correlated inversely with BMI in the range 20–27 [223]. In that study, BPND in other brain regions correlated with the BOLD signal response in orbitofrontal cortex upon viewing palatable food. In a group of lean subjects, the [11C]Caf (**8**) BPND at baseline in widespread brain regions correlated with BOLD responses to viewing palatable food [224], suggesting that low basal occupancy increases the response to cues. Interestingly, exercise increased or decreased thalamic μOR binding in these subjects; the direction of this change correlated with the individual BOLD signal in the contrast between viewing palatable and non-palatable food. This draws attention to individual differences in effects of exercise on the hedonic response to food, which may have some bearing on the relationship between exercise and weight loss, with the caveat that only intense exercise may significantly engage opioid transmission, as claimed above. Other studies show widespread reductions in μOR availability in frontolimbic regions after high intensity aerobic exercise, in correlation with negative affect. In contrast, mean binding was unaltered after moderate-intensity exercise, although there was some positive association with euphoria [225]; too much of a good thing spoils runner's high, it seems.

#### **4. Conclusions and Outlook**

The past decades have seen extraordinary progress in the development of ligands for PET studies of ORs. Early radiopharmaceutical research focused on studies with the antagonist [11C]DPN (**15**) and the μOR-selective agonist [11C]Caf (**8**), and the great preponderance of human PET studies have employed these and closely related tracers. While studies with non-selective tracers reveal the

composite of OR binding sites, specific tracers may be more indicative of physiological changes in disease states. Attaining high molar activity is of great importance in PET studies with [11C]Caf (**8**) and other highly potent agonist ligands; fortunately, most tracers described in this review have molar activities of at least 50 GBq/μmol, corresponding to an injected mass of about 1 μg of the drug. This is hardly a relevant dose in the case of antagonist ligands, and would give a comfortable 100-fold margin of safety with the potent μOR agonist [11C]Caf (**8**).

The μOR ligands have the useful property of binding in competition with endogenous opioid peptides, such that changes in the uptake in PET studies can reveal altered endogenous opioid release under various physiological conditions. This model has been particularly useful in studies of pain pathways, which largely involve μORs in telencephalon, and in some pharmacological or behavioral activation studies. However, the simple competition model may be inadequate to account for all observations. Thus, one of the [11C]Caf (**8**) studies noted above reports widespread reductions in μOR availability after smoking [207], despite the 30–3000-fold lower affinity of endogenous opioid peptides at μOR. This would seem to imply an implausibly enormous increase in peptide release to effect such changes by competition alone.

Pain studies have so far dominated the field of clinical PET research with OR-ligands, with relatively few reports on other models or conditions, as summarized in Table 2. For example, there in only one PET study of opioid receptors in schizophrenia, and only scant documentation in depression, or for that matter, in a range of common neurological disorders. In several human diseases noted above, the OR binding may be only 10% higher or lower than in the control group; while these small differences can have a large effect size, it is perhaps difficult to argue that such small differences can be causative of complex disorders or symptoms.



Addiction research using PET studies of ORs are so far mostly confined to alcohol, cocaine, and nicotine abuse and (strangely, perhaps), opioid addiction has hardly been a research theme, other than in a few occupancy studies. Since antagonists are relatively safe at doses provoking high occupancy (*viz* 50 mg naloxone for opioid overdose), we suppose that the Bmax of ORs might be determinable in relation to opioid addiction and withdrawal by conducting serial PET studies over a range of molar activity, even in the presence of significant agonist occupancy. Indeed, chronic morphine was reported 45 years ago to increase the abundance of [3H]naloxone binding sites in rat brain [226], but no such studies are reported in human opioid users, despite the catastrophe of the current opioid addiction epidemic. This kind of information might help to understand better the correlates of addiction and relapse. In addition, genetic studies of dopaminergic and opioid systems in relation to addiction [227], in conjunction with molecular imaging studies, could help to establish better the risk factors for opioid addiction. Endomorphins and other novel opioid petpides may present new avenues for obtaining opioid analgesia [228], while moderating the risk of "iatrogenic opioid addiction". The development of PET tracers with good binding properties *in vivo* and high selectivity for ORs other than the μ-type has accelerated in the past decade. However, there remain relatively few clinical molecular imaging studies of these important targets. Thus, developments in radioligand chemistry have for the presence to read for the present presence outpaced clinical PET imaging, a state of affairs that could enable and motivate a broad range of studies focusing on non-μORs over the coming decades. Just for example, κORs have an established role in the reinstatement of stress induced drug use in experimental animals, i.e. nicotine use [229], and very recent results indicate a relationship between κORs and stress-induced binge cocaine use [197].

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**



#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Current Landscape and Emerging Fields of PET Imaging in Patients with Brain Tumors**

**Jan-Michael Werner 1, Philipp Lohmann 2, Gereon R. Fink 1,2, Karl-Josef Langen 2,3 and Norbert Galldiks 1,2,\*,**†


Academic Editor: Peter Brust

Received: 28 February 2020; Accepted: 20 March 2020; Published: 24 March 2020

**Abstract:** The number of positron-emission tomography (PET) tracers used to evaluate patients with brain tumors has increased substantially over the last years. For the management of patients with brain tumors, the most important indications are the delineation of tumor extent (e.g., for planning of resection or radiotherapy), the assessment of treatment response to systemic treatment options such as alkylating chemotherapy, and the differentiation of treatment-related changes (e.g., pseudoprogression or radiation necrosis) from tumor progression. Furthermore, newer PET imaging approaches aim to address the need for noninvasive assessment of tumoral immune cell infiltration and response to immunotherapies (e.g., T-cell imaging). This review summarizes the clinical value of the landscape of tracers that have been used in recent years for the above-mentioned indications and also provides an overview of promising newer tracers for this group of patients.

**Keywords:** amino acid; FET; FACBC; FDOPA; immunoPET; molecular imaging; glioma; brain metastases

#### **1. Introduction**

For the management of patients with brain tumors, clinicians frequently need to rely on imaging information obtained from anatomical magnetic resonance imaging (MRI) before, during, and after the treatment. While contrast-enhanced MRI is of paramount value in neuro-oncology, its specificity for neoplastic tissue is low, and changes of the blood-brain barrier permeability as indicated by contrast enhancement are not limited to tumor tissue [1–7]. Nevertheless, precise delineation of tumor extent, including non-enhancing tumor subregions, is decisive for several diagnostic and therapeutic steps (e.g., planning of biopsy, surgery, or radiotherapy) [7,8]. Following radioand/or chemotherapy, neurooncologists often encounter treatment-related changes. Some of these, e.g., pseudoprogression, are difficult to differentiate from actual tumor progression with conventional MRI alone [6–12]. Pseudoprogression describes a phenomenon characterized by an increase of contrast enhancement without clinical deterioration, and which disappears again over time without any treatment change [3,9,13–16]. Such treatment-related changes may occur early (in the case of pseudoprogression typically within the first 12 weeks after chemoradiation completion) or late (several months or even years after radiotherapy in the case 31 of radiation necrosis) [7,8,17].

Furthermore, surrogates of treatment response or progression obtained from MRI (e.g., a decrease of contrast enhancement or the fluid-attenuated inversion recovery (FLAIR) signal hyperintensity) may be unspecific. They can be influenced by inflammation, infarction, and reactive changes after surgery [2,6,12,18]. If treatment-related changes remain unidentified, an effective treatment may be erroneously terminated prematurely. The latter may also harm survival and mislead study results evaluating novel treatment approaches for tumor relapse [19].

To overcome these diagnostic challenges, imaging techniques with higher diagnostic accuracy than conventional MRI offering more than just anatomical information are needed. Apart from advanced MRI techniques, positron-emission tomography (PET) imaging has been evaluated over the past decades. It has been shown that PET imaging offers additional value in neuro-oncology since it enables the non-invasive evaluation of molecular and metabolic features of brain tumors. PET, therefore, is of great value for the indications mentioned above, which are of particular clinical interest [7,8,10,20]. Consequently, the PET task force of the Response Assessment in Neuro-Oncology (RANO) working group highlighted the additional clinical value of PET imaging using amino acid tracers compared to anatomical MRI. Accordingly, its widespread clinical use was recommended in patients with glioma and brain metastases [17,21].

The continuously growing landscape of PET tracers enables the evaluation of many biochemical processes in patients with brain tumors. With the advent of newer treatment options in neuro-oncology, in particular, targeted therapy and various immunotherapy options, the needs for additional information derived from neuroimaging in terms of characterization of the tumor environment, the evaluation of tumoral drug accumulation, immune cell infiltration, and the diagnosis of treatment-related changes following these newer treatment options are steadily increasing. Some of these requirements may be met by the existing landscape of well-established PET tracers, while others can be addressed by newer ones [22,23].

This review summarizes the value of PET tracers that have been used in brain tumors in recent years for the most relevant clinical indications. Furthermore, more unique but promising PET tracers are summarized and discussed.

#### **2. Methods**

A PubMed search using the terms "PET", "positron", "tracer", "glioma", "brain metastases", "FDG", "amino acid", "methionine", "FET", "FDOPA", "FACBC", "AMT", "TSPO", "GE-180", "FLT", "FAZA", "EGFR", "VEGF", "immunoPET", "isocitrate dehydrogenase", "radiotherapy", "T-cell imaging", "reporter gene", "radiation necrosis", "pseudoprogression", "tumor extent", "response assessment", "treatment-related changes", and combinations thereof was performed until January 2020. The PET tracers were evaluated regarding their clinical value for the delineation of tumor extent, diagnosis of treatment-related changes, the assessment of treatment response (Table 1), and according to the information provided by newer PET probes (Table 2).



 = highincreased accuracy when using dynamic [18F]FET PET; = in enhancing and non-enhancing tumors; = in patients undergoing antiangiogenic treatment with bevacizumab; [11C]AMT = α-[11C]-methyl-L-tryptophan; [18F]FACB = anti-1-amino-3-[18F]fluorocyclobutane-1-carboxylic acid; [18F]FAZA = [18F]flouroazomycin arabinoside; [18F]FDG = [18F]-2-fluoro-2-deoxy-D-glucose; [18F]FDOPA = 3,4-dihydroxy-6-[18F]fluoro-L-phenylalanine; [18F]FET = O-(2-[18F]fluoroethyl)-L-tyrosine; [18F]FLT = 3--deoxy-3--[18F]flurothymidine; [18F]FMISO = [18F]fluoromisonidazole; [15O]H2O = radiolabeled water; [11C]MET = [11C]methyl-L-methionine; n.a. = only preliminary or no data available.

++



receptor; [18F]FHBG = 9-[4-[18F]fluoro-3-(hydroxymethyl)butyl]guanine; HER2 = human epidermal growth factor receptor 2; HSV1-tk = herpes simplex virus type 1 thymidine kinase; IDH = isocitrate dehydrogenase-1 or -2; PD-1 = programmed cell death receptor-1; PD-L1 = programmed cell death protein ligand 1; TKI = tyrosine kinase inhibitor.

#### **3. Current Landscape of PET Imaging**

PET allows targeting metabolic and molecular processes in patients with brain tumors relevant to diagnosis, treatment, and prognosis that cannot be assessed with anatomic computed tomography (CT) or MR imaging. A variety of PET tracers have been evaluated predominantly in glioma patients or patients with brain metastases with the main focus on glucose metabolism, amino acid transport, proliferation, hypoxia, blood flow, or angiogenesis. This section will provide an overview of PET tracers for brain tumors that have been evaluated in human subjects in the last years, especially for those above mentioned highly relevant indications in clinical routine. An overview is presented in Table 1.

#### *3.1. PET Imaging of Glucose Metabolism*

[ 18F]-2-Fluoro-2-deoxy-D-glucose ([18F]FDG) is the most widespread PET tracer in nuclear medicine. In neoplastic tissue, the uptake of [18F]FDG reflects the increased expression of glucose transporters and hexokinase. The latter enzyme phosphorylates glucose and [18F]FDG. In the central nervous system, the physiologically high and varying uptake of [18F]FDG in healthy brain parenchyma hampers the accurate delineation of the brain tumor. This limits the diagnostic accuracy for the correct identification of treatment-related changes and assessment of treatment response in gliomas and brain metastases [17,21]. It has repeatedly been shown that the diagnostic accuracy of [18F]FDG regarding the differentiation of radiation-induced changes from glioma and brain metastases recurrence is inferior to other imaging modalities, including advanced MRI and amino acid PET [17,24–26]. However, [ 18F]FDG PET seems to be of value for the delineation of tumor extent and assessment of treatment response in patients with primary central nervous system (CNS) lymphoma [27–30].

#### *3.2. PET Using Amino Acid PET Tracers*

Radiolabeled amino acid tracers (Figure 1) are of great interest in brain tumor imaging because of the high tumor-to-brain contrast based on the relatively high specificity for neoplastic tissue and the low uptake in healthy brain tissue [7,8,21,31–33].

**Figure 1.** Chemical structure of radiolabeled amino acids.

#### 3.2.1. Uptake Mechanisms of Amino Acid PET Tracers

The uptake of the amino acid tracers O-(2-[18F]fluoroethyl)-L-tyrosine ([18F]FET), [ 11C]methyl-L-methionine ([11C]MET), and 3,4-dihydroxy-6-[18F]fluoro-L-phenylalanine ([18F]FDOPA) is mainly based on the increased expression of large neutral amino acid transporters of the l-type (LAT) in gliomas and brain metastases (i.e., subtypes LAT1 and LAT2) [7,34–37]. Moreover, LAT1 overexpression correlates with malignant phenotypes and proliferation of gliomas. It is associated with glioma angiogenesis [38,39]. A critical consideration for the practical application of [11C]MET compared to [18F]FET or [18F]FDOPA is the half-life of the [11C]-isotope ([18F] 110 vs. [11C] 20 min) [40,41], which allows the transport of [18F]FET and [18F]FDOPA to PET facilities. In contrast, the use of [ 11C]MET necessitates an on-site cyclotron. In many European centers, this logistical disadvantage has led to the replacement of [11C]MET predominantly by [18F]FET [7]. When using [18F]FDOPA, the physiological uptake in the striatum may hamper the evaluation of tumor extent [7,42].

The L-tryptophan analogue α-[11C]-methyl-L-tryptophan ([11C]AMT) is another radiolabeled amino acid with uptake via the LAT system. Additionally, [11C]AMT uptake is mediated via the kynurenine pathway and has a rate-limiting enzyme indoleamine 2,3-dioxygenase [43]. Indoleamine 2,3-dioxygenase is upregulated in various cancers including gliomas [44], which prompted the use of [ 11C]AMT PET in patients with brain tumors [45].

Other tracers such as the synthetic amino acid analog anti-1-amino-3-[18F]fluorocyclobutane-1-carboxylic acid ([18F]FACBC or [18F]fluciclovine) are also LAT-mediated but use additionally the alanine, serine, and cysteine transporter 2, which is upregulated in many human cancers [46–49].

#### 3.2.2. Value of Amino Acid PET Tracers for Brain Tumor Patients

For the planning of diagnostic and therapeutic procedures, the precise delineation of tumor spread is essential. For example, the tumor extent as assessed by amino acid PET provides valuable information for planning stereotactic biopsies, resection, and radiotherapy [1,50–52]. For [11C]MET and [ 18F]FET, it has been shown that the delineation of tumor extent, particularly in non-enhancing gliomas, can be assessed with high accuracy using amino acid PET [51,53]. Preliminary data suggest that newer tracers such as [18F]FACBC PET are also helpful in identifying metabolically active and non-contrast enhancing tumor regions in glioma patients [54–56]. Moreover, it has been shown that in the majority of cases, the metabolically active tumor burden as assessed by amino acid PET extends considerably beyond the volume of MRI contrast enhancement, which is of significant relevance for subsequent treatment planning [1,7,53,57]. Regarding the comparability of PET tracers for this indication, [11C]MET, [ 18F]FET, and [18F]FDOPA seem to be equally informative [58–61]. Nevertheless, it has to be pointed out that 20–30% of grade II gliomas, according to the World Health Organization (WHO) classification of tumors of the central nervous system [62,63], show no amino acid uptake [64–66]. A negative amino acid PET, therefore, does not exclude glioma [8].

For the differentiation of treatment-related changes from tumor relapse, amino acid PET also provides valuable diagnostic information. Using [18F]FET or [18F]FDOPA, especially the differentiation of radiation injury from tumor relapse in glioma patients, as well as in patients with brain metastases, can be obtained with a relatively high diagnostic accuracy between 80–90% (Figure 2) [11,26,67–80]. Importantly, in glioma patients, parameters derived from dynamic [18F]FET PET acquisition may further increase the diagnostic accuracy [67,74–76,78]. This has also been demonstrated in patients with brain metastases who underwent radiosurgery for brain metastases treatment [80,81]. The diagnostic accuracy of [11C]MET PET regarding this clinical question is slightly lower (approximately 75%) [8,82,83], which is most probably related to the higher affinity of [11C]MET for inflammation [84]. First PET studies using [11C]AMT or [18F]FACBC suggest that these tracers may also be of value for the differentiation of radiation injury from glioma progression [85,86].

**Figure 2.** Radiation necrosis and chronic inflammation in a patient with brain metastases of a B-Raf proto-oncogene (BRAF)-mutated malignant melanoma who had been treated with whole-brain radiation therapy combined with concurrent dabrafenib plus trametinib. Twenty-four months later, the contrast-enhanced magnetic resonance imaging (MRI) indicates a recurrence of the brain metastases (left panel), whereas the O-(2-[18F]fluoroethyl)-L-tyrosine ([18F]FET) positron-emission tomography (PET) shows only insignificant metabolic activity and is consistent with the findings of treatment-related MRI changes. Neuropathological findings (right panel) after stereotactic biopsy show signs of radiation necrosis as well as considerable T-cell infiltration. (**A**) Hyaline, eosinophilic necrosis with evidence of a necrotic vessel wall (arrowhead). (**B**) Vital brain parenchyma besides necrosis with activated microglia cells (arrowhead), and blood vessels with lymphocyte infiltrates (arrows) without evidence of tumor cells (inserted box). (**C**) Adjacent to inflamed blood vessels (arrows), a resorption of necroses by macrophages (block arrows) as well as activated microglia cells (arrowheads) and astrocytes in the brain parenchyma (inserted box). (**D**) The main population of intra- and perivascular T-cell infiltrates are CD3<sup>+</sup> (arrow), but also CD4<sup>+</sup> (inserted box left) and CD8<sup>+</sup> (inserted box right) T-cells contribute to the infiltrates (modified from Galldiks et al. [10], with permission from Oxford University Press).

The advent of immunotherapy using immune checkpoint inhibitors and targeted therapy has improved the survival of cancer patients, particularly in melanoma and lung cancer. Recent trials suggest that patients with brain metastases from these tumor entities may also benefit from these agents alone or in combination [87]. Regarding patients with brain metastases treated with checkpoint inhibitors or targeted therapy (frequently combined with radiotherapy), initial data indicate that amino acid PET may provide valuable information for differentiating relapse from equivocal MRI findings related to immunotherapy-induced inflammation [10,88].

Recently, a variety of experimental treatment options has been introduced for treating patients with high-grade glioma. [18F]FET PET was shown to differentiate benign MRI findings related to these experimental therapies, e.g., immunotherapy with dendritic cell vaccination or targeted therapy with regorafenib, from tumor relapse [89,90]. However, the number of patients treated with these therapies and monitored with [18F]FET PET is still small, and the results should be interpreted with caution.

The assessment of the response to a particular neurooncological treatment is of clinical relevance since treatment decisions can be negatively affected by treatment-related changes. The accurate assessment of response helps both to discontinue an ineffective treatment option as early as possible and to prevent an effective treatment from being erroneously terminated prematurely with a potentially harmful influence on survival. Furthermore, the evaluation of response also helps to avoid possible treatment side effects, e.g., bone marrow depression or fatigue, and, therefore, to maintain or even improve life-quality. It has been shown in glioma patients that the assessment of response to alkylating chemotherapy (i.e., temozolomide or lomustine) using [11C]MET or [18F]FET PET provides valuable additional information compared to contrast-enhanced MRI. Importantly, metabolic PET responders (i.e., patients with a decrease of tumor-to-brain ratios or metabolically active tumor volumes at follow-up relative to baseline imaging) had a significantly longer survival than metabolic non-responders [91–98].

Following antiangiogenic therapy using bevacizumab, the use of reduced contrast enhancement as a surrogate marker for treatment response is not optimal due to a phenomenon called pseudoresponse [9]. Pseudoresponse describes a decrease of contrast enhancement related to a rapid restoration of the blood-brain barrier by antiangiogenic drugs [9]. However, a clinical benefit is not infrequently lacking in patients with an impressive radiological response (pseudoresponse). [18F]FET and [18F]FDOPA PET may provide valuable information regarding the identification of pseudoresponse [99–102]. Moreover, [ 18F]FDOPA and [18F]FET PET were also able to predict a favorable clinical outcome in bevacizumab responders [101–104].

After chemoradiation completion, newly diagnosed glioblastoma patients can be treated with tumor-treating fields therapy in addition to adjuvant temozolomide chemotherapy [105]. Initial studies suggest that amino acid PET might identify responding patients undergoing tumor-treating fields therapy, but it has to be considered that the response can also be related to the concurrently applied chemotherapy or delayed chemoradiation effects [106,107].

#### *3.3. PET Imaging of the Mitochondrial Translocator Protein*

PET ligands targeting the 18 kDa mitochondrial translocator protein (TSPO), located at the outer mitochondrial membrane and formerly known as the peripheral benzodiazepine receptor, are also of interest in neuro-oncology [108,109]. TSPO is associated with neuroinflammation due to its expression in activated microglia, endothelial cells, and infiltrating macrophages [109]. The PET ligand [ 11C]PK11195 was one of the first ligands evaluated for TSPO expression in glioma patients [110–112].

The recently introduced TSPO ligand GE-180 labeled with [18F] offers an increased binding specificity and was tested in patients with gliomas [113] and neuroinflammatory diseases such as multiple sclerosis [114–116]. Regarding the delineation of glioma extent, it has been demonstrated that the [18F]GE-180 uptake volume is significantly larger than the volume of contrast enhancement [113,117]. However, when comparing [18F]FET with [18F]GE-180 uptake volumes intraindividually in terms of spatial distribution, the overlap is only moderate (Dice similarity coefficient, 0.55) despite comparable tumor volumes [117]. These differences might help to characterize glioma heterogeneity and warrant further studies with spatial correlation of imaging findings of [18F]FET uptake to [18F]GE-180 uptake with neuropathology.

#### *3.4. PET Imaging of Cellular Proliferation*

The radiolabeled nucleoside 3- -deoxy-3- -[18F]fluorothymidine ([18F]FLT) is a pyrimidine analogue. It is used to evaluate cellular proliferation because of its rapid incorporation into newly synthesized DNA [118]. [18F]FLT is trapped intracellularly after phosphorylation by the thymidine kinase-1, a cytoplasmatic enzyme expressed during cell proliferation [46,119]. However, the requirement of a disrupted blood-brain barrier for [18F]FLT uptake may limit its diagnostic use [2,120]. For example, in terms of tumor detection and delineation, [18F]FLT PET was less sensitive than [11C]MET PET to detect WHO grade II gliomas, which usually show no contrast enhancement [121]. Furthermore, a meta-analysis evaluating the value of [18F]FLT PET for the diagnosis of glioma recurrence based

on approximately 800 patients showed no superiority of [18F]FLT (pooled sensitivity, 82%; pooled specificity, 76%) compared to [18F]FDG (pooled sensitivity, 78%; pooled specificity, 77%) [122].

On the other hand, [18F]FLT PET seems to be useful for the assessment of response to antiangiogenic therapy with bevacizumab in patients with recurrent malignant glioma. [18F]FLT PET was able to identify a reduction of proliferative activity in responding patients with favorable outcome as an indicator for response compared to metabolic non-responders [123–125]. Furthermore, [18F]FLT PET was used in patients with malignant melanoma brain metastases treated with targeted therapy or immunotherapy using checkpoint inhibitors [126]. In that study, responding patients showed a clearer reduction of proliferative activity as assessed by [18F]FLT PET than the decrease of contrast enhancement on standard MRI.

#### *3.5. PET Imaging of Tumor Hypoxia*

Hypoxia is a key factor in treatment outcome in various cancers, including glioma. It has been shown that hypoxia is associated with tumor persistence and resistance to cancer treatment [127]. To further evaluate this phenomenon using PET, the tracer [18F]fluoromisonidazole ([18F]FMISO) has been developed, which is trapped in hypoxic but viable cells [46,128,129]. It has been demonstrated in glioblastoma patients that [18F]FMISO PET delineates additional hypoxic tumor subregions which exceed the contrast-enhancing tumor parts, indicating that hypoxia may induce peripheral tumor growth [130]. A subsequent study showed that the metabolically active tumor volume in [11C]MET PET strongly correlated with the hypoxic volume defined by [18F]FMISO [131]. Importantly, the tumor area on [11C]MET PET exceeded the area of the contrast enhancement on MRI in the range of 20–30%.

More recently, [18F]FMISO PET has been used for monitoring the effects of antiangiogenic therapy with bevacizumab in patients with recurrent malignant glioma [132,133]. It was shown that patients who had a response in both contrast-enhanced MRI and [18F]FMISO PET had significantly longer survival than patients who responded on MRI only [133].

[ 18F]-labeled flouroazomycin arabinoside ([18F]FAZA) may be a promising alternative to [ 18F]FMISO offering an improved tumor-to-background ratio due to faster blood clearance [134,135]. Preliminary data in glioblastoma patients suggest that [18F]FAZA PET might be of value for radiotherapy response assessment [136].

#### *3.6. PET Imaging of Tumor Perfusion*

The evaluation of regional cerebral blood flow (rCBF) allows identifying brain tumors with high vascularization. rCBF can be measured by PET using [15O]-labeled water. However, [15O]H2O requires an on-site cyclotron because of its very short half-life (2 min) [137]. The development and easy accessibility of perfusion-weighted CT and MRI has led to various studies evaluating brain tumors outnumbering [15O]H2O PET studies by far. In direct comparisons, it has been shown that rCBF values differ considerably between CT- and MRI-based perfusion measurements and [15O]H2O PET [138,139].

In patients with malignant gliomas undergoing chemoradiation with nitrosoureas, a reduction of rCBF was evaluated using [15O]H2O PET, but data on subsequent survival as an indicator for treatment response are lacking [140,141].

#### *3.7. PET Imaging of Angiogenesis*

The vascular endothelial growth factor (VEGF) is overexpressed by most tumors, including brain tumors, and an important trigger for neovascularization [142]. Bevacizumab is a recombinant humanized monoclonal antibody against VEGF and appears to prolong progression-free survival and decrease steroid usage in patients with malignant glioma [142,143]. Based on these properties, [ 89Zr]-labeled bevacizumab PET imaging has been evaluated for tumoral drug accumulation in pediatric patients with diffuse intrinsic pontine glioma [144]. That study demonstrated intertumoral heterogeneity of drug accumulation and may aid in selecting those patients with the greatest chance of benefit from bevacizumab [144,145]. Other studies have used [64Cu]-labeled conjugates of VEGF

with DOTA (1,4,7,10-tetra-azacylododecane *N*,*N*- ,*N*--,*N*----tetraacetic acid) for PET imaging in animal models [146–148]. Similarly, these approaches suggest the clinical value for the identification of patients who will benefit from anti-VEGF therapy.

In adult glioblastoma patients, it has been demonstrated that accumulation of [123I]VEGF in the tumor region of glioblastomas can be assessed using single photon emission computed tomography (SPECT) imaging [149]. Importantly, high uptake of [123I]VEGF was able to identify glioblastoma patients with a poor clinical outcome.

#### **4. Emerging Fields of PET Imaging**

Currently, blockade of immune checkpoints and other immunotherapy options (i.e., vaccination strategies, oncolytic virus approaches, cell-based immunotherapy such as chimeric antigen receptor T-cells (CAR-T cells) are under evaluation in patients with brain cancer including glioma and brain metastases. Furthermore, recent study results suggest that newer-generation targeted therapies are a promising treatment option, especially in a subset of patients with brain metastases [150]. As a new treatment option, mutations of the isocitrate dehydrogenase (*IDH*) gene, which frequently occur in WHO grade II and III gliomas, have also gained interest as a potential treatment target [151,152]. All these promising treatment options impose new demands on brain imaging (e.g., imaging of immune reactions in the brain). The current body of literature suggests that PET has the potential to adapt to these needs. An overview is presented in Table 2.

#### *4.1. PET Imaging of the Epidermal Growth Factor Receptor Family*

Both the epidermal growth factor receptor (EGFR) and the human epidermal growth factor receptor 2 (HER2) are transmembrane protein receptors and belong to the EGFR family. These receptors are targets for various growth factors that mediate various cellular processes such as differentiation or proliferation. In clinical oncology, various gene mutations may lead to overexpression of these proteins and are associated with the development of a variety of cancers. Importantly, these mutations also play a significant role in various treatment options, including tyrosine kinase inhibitors and monoclonal antibodies targeting EGFR, HER2, or both [153], as well as for imaging. Especially the evaluation of a response to these targeted therapy options in patients with brain metastases (from melanoma, lung, or breast cancer) as well as in glioma patients is the indication with the highest clinical potential for EGFR- and HER2-targeted PET [154,155]. In recent years, radiolabeled EGFR and HER2 antibodies, as well as tyrosine kinase inhibitors, have been used as PET imaging agents.

For identifying EGFR overexpression, PET ligands such as [11C]erlotinib, [11C]PD153035, and [ 89Zr]Zr-DFO-nimotuzumab have been used [156–158]. The most relevant PET tracers for imaging of HER2 overexpression are [64Cu]DOTA-trastuzumab and [89Zr]pertuzumab (Figure 3) [159,160].

#### *4.2. Immuno-Imaging: Immuno-PET and Imaging of T-Cells*

In recent years, immunotherapy with antibodies directed against immune checkpoints such as the cytotoxic T-lymphocyte antigen-4 (CTLA-4; e.g., ipilimumab), the programmed cell death receptor-1 (PD-1; e.g., pembrolizumab, nivolumab), or the programmed cell death protein ligand 1 (PD-L1; e.g., atezolizumab) have gained paramount importance in clinical oncology and neuro-oncology. However, the efficacy and responsiveness of these agents may vary considerably among different cancer types and across individuals. Biomarkers obtained from tumor tissue, such as PD-1 and PD-L1 expression, can help to select patients. However, these tissue biomarkers are limited, and some patients show no response even if the target is present [161]. Therefore, the significance of PET for predicting response to immunotherapy and patient selection increases. Among other methods, immuno-PET combines antibodies or antibody fragments with a radionuclide and takes advantage of the specificity and affinity of antibodies and the sensitivity of PET [162]. Generally, targets for immuno-PET can be T-cell markers (e.g., CD4+, CD8+), immune checkpoints (e.g., CTLA-4, PD-1, PD-L1), or biomarkers of the immune response (e.g., interferon-γ, interleukin-2) [23].

**Figure 3.** [ 64Cu]-DOTA-trastuzumab positron-emission tomography (PET) and contrast-enhanced magnetic resonance imaging (MRI) performed one day after initiation of treatment with trastuzumab in a patient with a human epidermal growth factor receptor 2 (HER2)-positive breast cancer with brain metastases. In single brain metastases, [64Cu]-DOTA-trastuzumab PET helps to improve lesion detection (arrow) (modified from Tamura et al. [160], with permission from the Society of Nuclear Medicine and Molecular Imaging).

First-in-human studies suggest that targeting PD-1 with [89Zr]nivolumab [163] or PD-L1 with [ 89Zr]atezolizumab [164] is useful as imaging biomarkers to non-invasively evaluate the expression of these immune checkpoints in patients with extra- and intracranial cancer [164,165]. Engineered target-binding proteins (adnectins) for PD-L1 ligands such as [18F]BMS-986192 are currently under evaluation [166].

Tumor-infiltrating T-cells (such as CD8+) play an essential role in the activation of immune cells in response to checkpoint inhibition [167]. Recently, it has been shown that a radiolabeled [ 89Zr]IAB22M2C has the potential to visualize CD8<sup>+</sup> T-cell-enriched tumor tissue [168]. The assessment of immune cells infiltrating tumors has also been investigated with PET using radiolabeled clofarabine (2-chloro-2- -deoxy-2- -[18F]fluoro-9-b-D-arabinofuranosyl-adenine; [18F]CFA) [169]. [18F]CFA is a substrate for the enzyme deoxy-cytidine kinase, which is overexpressed in immune cells such as CD8<sup>+</sup> T-cells [169]. Importantly, [18F]CFA PET has shown a great clinical potential to localize and quantify immune responses in glioblastoma patients undergoing dendritic cell vaccination treatment combined with immune checkpoint blockade (Figure 4) [169].

Another interesting approach is the transfection of immune cells with a reporter gene that encodes a protein that can be specifically targeted by a radiolabeled reporter probe [170]. Imaging of reporter gene expression of cells transfected with the herpes simplex virus type 1 thymidine kinase reporter gene has been demonstrated using 9-[4-[18F]fluoro-3-(hydroxymethyl)butyl]guanine ([18F]FHBG) [171]. Moreover, this technique has the potential to image recently introduced cell-based therapies with CAR-T cells [23]. Accordingly, a study with recurrent high-grade glioma patients suggested that [ 18F]FHBG PET can detect reporter gene expression in CAR-engineered cytotoxic T-lymphocytes [172].

#### *4.3. PET Imaging of Isocitrate Dehydrogenase Mutations*

In patients with malignant glioma, the modest outcome improvement following both standard therapy (i.e., chemoradiation with temozolomide) and newer treatment options (e.g., tumor-treating fields) has prompted various efforts to identify molecules that are fundamental to regulate tumor progression and provide additional options for personalized therapy in this group of patients.

**Figure 4.** Detection of immune response in a patient with recurrent glioblastoma using 2-chloro-2- -deoxy-2- -[18F]fluoro-9-b-D-arabinofuranosyl-adenine ([18F]CFA) positron-emission tomography (PET) and advanced magnetic resonance imaging (MRI) before (upper panel) and after treatment with dendritic cell vaccination and programmed cell death receptor-1 (PD-1) blockade using pembrolizumab (lower panel). Following treatment, [18F]CFA uptake is considerably increased, indicating an immune-cell infiltration, and helps distinguishing tumor progression from inflammation (modified from Antonios et al. [169], with permission from the National Academy of Sciences).

Accordingly, the enzyme isocitrate dehydrogenase (IDH) has gained interest as a potential target. IDH is an enzyme of the Krebs cycle, catalyzing the oxidative decarboxylation of isocitrate to alpha-ketoglutarate. Mutations in the *IDH1* and *IDH2* gene, frequently occurring in WHO grade II or III astrocytomas and oligodendrogliomas, result in a significant increase of the oncometabolite 2-hydroxyglutarate (2-HG) [173,174]. In cells with IDH mutant enzymes, the accumulation of 2-HG alters several downstream cellular activities, causing epigenetic dysregulation and, consequently, a block in cellular differentiation, leading to oncogenesis [175].

Therefore, mutant IDH proteins are highly attractive targets for inhibitory drugs. In glioma patients, selective oral IDH inhibitors of IDH1 (i.e., ivosidenib, BAY-1436032), pan-IDH1/2 (i.e., AG-881), and vaccination strategies targeting the IDH1R132H mutation are currently under clinical evaluation. Initial results predominantly from phase-1 studies are promising and suggest that these inhibitors are safe and have antitumoral activity [151,152,176]. Although immunohistochemistry and genomic sequencing are the methods of choice for the detection of an IDH mutation, these techniques are invasive and are not appropriate for treatment monitoring, which requires continual assessment. Furthermore, the use of magnetic resonance spectroscopy (MRS) to non-invasively evaluate 2-HG is technically challenging, and may be false-positive in 20% of cases [177,178].

Newer PET probes for imaging mutant IDH expression in gliomas may be an alternative imaging method. Recent radiochemical developments suggest that triazinediamine or butyl-phenyl sulfonamide analogs labeled with [18F] are promising candidate radiotracers for noninvasive PET imaging of IDH mutations in gliomas [179,180]. Furthermore, a [18F]-labeled IDH1 inhibitor (AGI-5198) has also been investigated [181]. Interestingly, Koyaso and colleagues demonstrated that [11C]acetate uptake in IDH mutant cells is significantly higher than in IDH wild-type cells because of metabolic trapping [182]. Taken together, further efforts to translate these promising approaches for IDH imaging into clinical use are warranted.

#### *4.4. PET-Based Theranostics*

The combination of therapeutics and diagnostics, also termed as theranostics, supports the concept of precision oncology. One PET-based theranostic approach is the peptide receptor radionuclide therapy (PRRT), in which overexpressed tumor-specific receptors are used as a therapeutic target. By exchanging the radionuclide used for diagnostic PET such as [68Ga] with a radiation source, typically *ß*-emitters like [177Lu] or [90Y], the same PET tracer can be used for therapy. Although there are currently no theranostic approaches clinically established for gliomas or brain metastases, there are promising concepts. For example in patients with glioblastoma, a potential target is the overexpressed chemokine receptor-4 (CXCR4) which is associated with a poor clinical outcome [183–185]. Visualization of CXCR4 expression using diagnostic PET with CXCR4-directed [68Ga]Pentixafor® has been demonstrated in glioblastoma patients [186].

Another potential target is the prostate-specific membrane antigen (PSMA) which may be overexpressed in prostate cancer and also in predominantly malignant gliomas. Diagnostic PET imaging of PSMA expression in patients with malignant glioma can be obtained using [68Ga]-labeled PSMA ligands [187–189]. The use of the theranostic agent [177Lu]-PSMA-617 has demonstrated favorable safety and efficacy in patients with advanced prostate cancer, indicating the potential to be also of value for patients with malignant glioma [190]. Nevertheless, the potential of theranostics needs to be further evaluated in patients with malignant gliomas.

#### **5. Discussion**

Anatomical MRI is currently the method of choice for neuroimaging of brain tumors, but PET complements this technique and provides important biological information that cannot be obtained from anatomical MRI alone. Currently, best-established PET tracers in neuro-oncology are radiolabeled amino acids targeting L-system transporters. However, a considerable number of other PET tracers have been developed for brain tumor patients and allow the evaluation of a wide range of biochemical processes. A variety of PET biomarkers offers the potential to play a clinically significant role for the monitoring of newer treatment options such as targeted therapy and immunotherapy, e.g., by providing an early assessment of response to these options, and for a more accurate differentiation of viable tumor from treatment-related changes. A major current shortcoming is the lack of large, prospective clinical trials in patients with both glioma and brain metastases for many of these PET tracers. Despite encouraging early study results in the field, it has to be further demonstrated that these tracers improve considerably patient management and outcome.

**Author Contributions:** Study design, writing of manuscript drafts: J.W. and N.G. Revising manuscript, approving final content of the manuscript: All. All authors have read and agreed to the published version of the manuscript. **Funding:** The Cologne Clinician Scientist-Program (CCSP) of the Deutsche Forschungsgemeinschaft (DFG, FI 773/15-1), Germany, supported this work.

**Conflicts of Interest:** Related to the present work, the authors disclosed no potential conflicts of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Approaches to PET Imaging of Glioblastoma**

**Lindsey R. Drake 1,2,\*, Ansel T. Hillmer 1,2,3,4 and Zhengxin Cai 1,2**


Academic Editor: Peter Brust

Received: 11 December 2019; Accepted: 23 January 2020; Published: 28 January 2020

**Abstract:** Glioblastoma multiforme (GBM) is the deadliest type of brain tumor, affecting approximately three in 100,000 adults annually. Positron emission tomography (PET) imaging provides an important non-invasive method of measuring biochemically specific targets at GBM lesions. These powerful data can characterize tumors, predict treatment effectiveness, and monitor treatment. This review will discuss the PET imaging agents that have already been evaluated in GBM patients so far, and new imaging targets with promise for future use. Previously used PET imaging agents include the tracers for markers of proliferation ([11C]methionine; [18F]fluoro-ethyl-L-tyrosine, [18F]Fluorodopa, [ 18F]fluoro-thymidine, and [18F]clofarabine), hypoxia sensing ([18F]FMISO, [18F]FET-NIM, [18F]EF5, [ 18F]HX4, and [64Cu]ATSM), and ligands for inflammation. As cancer therapeutics evolve toward personalized medicine and therapies centered on tumor biomarkers, the development of complimentary selective PET agents can dramatically enhance these efforts. Newer biomarkers for GBM PET imaging are discussed, with some already in use for PET imaging other cancers and neurological disorders. These targets include Sigma 1, Sigma 2, programmed death ligand 1, poly-ADP-ribose polymerase, and isocitrate dehydrogenase. For GBM, these imaging agents come with additional considerations such as blood–brain barrier penetration, quantitative modeling approaches, and nonspecific binding.

**Keywords:** PET imaging; GBM; biomarkers; Sigma 1; Sigma 2; PD-L1; PARP; IDH

#### **1. Introduction**

Glioblastoma Multiforme (GBM) is a fast growing, invasive brain tumor that typically results in death in the first 15 months after diagnosis [1]. It develops from glial cells, astrocytes or oligodendrocytes, and can evolve from lower-grade tumors or de novo. Previously, GBM was characterized as 'grade IV' astrocytoma. Recently, the World Health Organization (WHO) updated the classification of brain tumors to include genotypic markers, building on the histological markers considered previously [2]. Glioblastoma can be classified by a single nucleotide polymorphism in the isocitrate dehydrogenase (IDH) gene as wild-type or mutant. Approximately 10% of glioblastomas are IDH-mutant [2]. IDH-mutant status weakly predicts long-term survival (over 3 years post diagnosis) [3]. GBM tumors are heterogenous in location (with 25%–43% incidence in frontal lobes), histopathology, and the tumor microenvironment [4]. The first line of treatment for GBM is surgery, followed by radiation and chemotherapy [1]. Temozolomide, a DNA alkylating agent is often used for chemotherapy. In 2015, the vascular endothelial growth factor inhibitor Bevacizumab was fast-tracked for use in GBM after demonstrating efficacy in shrinking or halting tumor growth. However, it has failed to show

improvement in overall survival [5]. Patients with GBMs have a very low survival rate with very few treatment options, making this a particularly acute health challenge.

Medical imaging provides critical information for diagnosing, staging, and monitoring the treatment of GBM. While formal diagnosis relies on histopathology and genetic markers for grading, structural magnetic resonance images (MRIs) are routinely acquired and can be used in guiding surgery. Additional structural MRI methods can accurately classify and grade tumors with high accuracy, though it has not been adopted yet as common practice [6]. Positron emission tomography (PET) imaging provides important complementary information to anatomical MRI data. In this functional type of imaging, biochemical information about the tumor and the tissue surrounding it can be measured non-invasively. GBMs typically are fast growing, giving an important role for specific PET radioligands to quantify proliferation. PET imaging is also uniquely positioned to identify ideal cases for targeted treatments and evaluate treatment progression.

This article provides an overview of the novel imaging tracers used in PET imaging of brain tumors. Discussion includes the strengths, limitations, and pitfalls of individual imaging biomarker strategies, and general challenges associated with PET imaging of brain tumors. We first provide a brief overview of established PET imaging biomarkers (glycolysis, amino acid metabolism, DNA replication, hypoxia, and inflammation), followed by newer imaging targets (Sigma 1/ 2, programmed death ligand 1, poly-ADP-ribose polymerase, and isocitrate dehydrogenase) with promise to image glioblastoma lesions. None of these biomarkers are unique to glioblastoma, though their presence has been found in resected brain tumors. This work concludes with important quantitative considerations for use of these imaging biomarkers in the evaluation and treatment of GBM patients.

#### **2. Overview of PET Imaging Agents for Brain Tumor**

#### *2.1. Sustained Proliferation Markers: Glycolysis, Amino Acid Transportation, and DNA Replication*

The classic approach to imaging tumors in general, and in application to GBM, has been to probe the functional necessities of proliferation. These necessities include glucose metabolism, protein synthesis, and DNA replication. From a biochemical prospective, these functions highlight the 'building block' small molecules that compose macromolecules: sugars, nucleotide bases, and amino acids.

Radionuclide-labeled forms of these building blocks have been employed to study these functions with PET imaging. The gold standard of most cancer imaging is [18F]FDG (**1**), a fluorine-18 glucose analogue. This radiotracer is actively taken up by the glucose transporter and participates in the first step of glucose metabolism (phosphorylation), then becomes trapped in the cell [7]. [18F]FDG PET allows for the functional imaging of glucose metabolism, a relative gold mine of information in most cancers. However, the brain has naturally high uptake of [18F]FDG, which complicates interpretation of GBM lesions near gray matter. Further efforts to image proliferation through the 'building block' strategy include neutral amino acid analogues ([11C]methionine (**2**); [18F]fluoro-ethyl-L-tyrosine (**3**); [ 18F]Fluorodopa (**4**)) and deoxynucleoside bases ([18F]fluoro-thymidine (**5**); and [18F]clofarabine (**6**)) (Figure 1).

[ 11C]Methionine ([11C]MET) was developed shortly after [18F]FDG [8,9]; it was considered a valuable tracer because methionine, an essential amino acid, can be used in protein synthesis. Both L-and D-isomers were synthesized, and no difference in accumulation was observed [10]. This lack of selectivity between stereoisomers indicated that [11C]MET was not being incorporated for protein synthesis, as only L-amino acids are incorporated into proteins. Despite this, [11C]MET was still used as an alternative to [18F]FDG PET imaging because it is more sensitive (see review [11]). [ 18F]fluoroethyl-tyrosine ([18F]FET) was developed as another amino acid based alternative to [ 18F]FDG [12]. Unlike [11C]MET, [18F]FET is a modified amino acid, which led to questions about the radiotracer's ability to be taken up into cells. In cellular experiments and tumor-bearing mice, it was found that [18F]FET is actively transported into cells [13]. The transporter responsible was later identified as l-amino acid transporter (LAT-1) [14]. [18F]FET PET combined with MRI can dramatically

improve glioma identification and tumor diagnosis [15]. In terms of the diagnostic performance, MRI alone yielded a 96% sensitivity and 53% specificity; the combined technique achieved 93% sensitivity and 94% specificity [15]. While MRI is the gold standard for diagnosis and tumor staging, [18F]FET PET imaging provides complementary information in the form of increased specificity.

**Figure 1.** Radiotracers for markers of cellular proliferation.

The nonessential amino acid and neurotransmitter dopamine has also been used as a PET tracer in the form of [18F]L-fluoro-dihydroxyphenylalanine ([18F]FDOPA), which becomes a dopamine analogue in vivo after decarboxylation. [18F]Fluoro-dopamine can then be further taken up into vesicles by VMAT, or metabolized by monoamine oxidases or catechol-*O*-methyltransferase. This gives [ 18F]FDOPA a particular utility in neuroendocrine tumors [16] in addition to the obvious neurological application. [18F]FDOPA imaging in pre-operative glioma patients has shown a significant correlation between WHO grade and the volume of MRI contrast enhancement, volume of T2 hyperintensity, and [18F]FDOPA uptake (as SUVmax ratio of tumor to normal tissue) [17]. From this 45 patient cohort, a multivariate Cox regression suggested that [18F]FDOPA PET and age were significant prognostic factors for overall survival [17]. A major limitation of [18F]FDOPA use is the radiosynthesis. One of the first described high yielding radiosynthesis involved electrophilic fluorination [18]. Electrophilic fluorination is not a desirable method for routine clinical production because of the hazardous nature of F2 gas, however, improved syntheses were recently published involving nucleophilic fluorination [19].

Concurrent with the use of amino acid analogues, the deoxynucleoside base derivative [ 18F]fluorothymidine ([18F]FLT) was developed. [18F]FLT is trapped in tissues after phosphorylation by thymidine kinase [20]. This trapping was thought to be due to accumulation into growing DNA chains, as fluorothymidine was originally developed to terminate DNA, and used in HIV therapy. In reality, less than 1% of [18F]FLT was found to be incorporated in DNA in cellular studies, though it correlated highly with [3H]thymidine uptake [21]. Despite this lack of accumulation in DNA, [18F]FLT imaging confirms cellular uptake and correlates to Ki-67 expression on the corresponding resected tumor tissue [22]. Thymidine kinase is a principal enzyme in the DNA salvage pathway, and is most active in G1 and S phases of the cell cycle [23]. Because of its lower uptake in normal brain than [ 18F]FDG, glioma patients underwent [18F]FLT imaging and a similar correlation was observed with proliferative tissue markers [24]. When [18F]FLT PET was used to monitor treatment in malignant glioma trials and compared to MRI responses, [18F]FLT-PET was more predictive of overall survival than MRI [25]. Despite this utility, [18F]FLT is less widely used because of nonspecific binding in

individuals [26], and because blood–brain barrier (BBB) penetration is limited, which limits its use specifically for glioma imaging.

An alternative method to [18F]FLT, which is phosphorylated by thymidine kinase, is to target deoxycytidine kinase (dCK). The activity of dCK, TK, and the other deoxyribonucleoside kinases are to provide an alternative pathway to de novo synthesis of DNA precursors [27]. Additionally, the activity of these kinases is critical for the activation of nucleoside analogues that are used for chemotherapy, like clofarabine [28]. Clofarabine is a chemotherapy used in pediatric patients with relapsed or refectory acute lymphoblastic leukemia. The radiolabeled [18F]clofarabine ([18F]CFA) has been evaluated in healthy humans and found to be BBB permeable [29]. Biodistribution of [18F]CFA showed uptake in lymph nodes, consistent with known dCK activity being required for T- and B-cell development [30]. In two patients with recurrent GBM, PET imaging with MRI was able to delineate specific regions of immune activity [31]. In one individual, the post-treatment PET-MRI scan demonstrated a 300% increase in immune cells in the tumor microenvironment, with the tumor volume remaining consistent. This combined technique has been shown in this preliminary data set to be useful in differentiating tumor progression from immune cell infiltration in a treatment monitoring scenario.

PET imaging of these classic markers of proliferation is useful to image GBM due to their simplistic design. However, changes in glucose metabolism, DNA replication, protein synthesis, and neurotransmitter homeostasis are not unique to cancers; they can describe a number of disease states. This generality makes these imaging approaches most appropriate for disease monitoring in cancer treatments like chemotherapy and targeted radiation. For further discussion of these tracers and their prognostic value in brain tumors, see a recent systematic review [32].

#### *2.2. Hypoxia-Sensing Tracers*

The first-line treatment strategy for gliomas is surgical resection if possible, followed by chemotherapy and targeted radiation [33]. Chemotherapy and radiation are therapies that hinge on cell death by damaging DNA and initiating apoptosis. However, some tumors are resistant to these strategies. For example, it has been long known that poorly oxygenated tissue, or hypoxic tissue, is less sensitive to radiation. Gray et al. confirmed in 1953 that X ray therapy was more effective in mice that were breathing oxygen at a higher pressures than normal atmosphere [34]. Hypoxia has become a recognized key feature of most solid tumors [35]. In a hypoxic tumor microenvironment, radiation therapy could be more effective at a higher dose; however, this requires an accurate identification of that cell population.

PET radiotracers that sense oxygen levels in cells can be used to visualize hypoxia (Figure 2). [ 18F]Fluoromisoinodazole ([18F]FMISO; **7**) contains a nitroimidazole which is reduced to RNO2 radical after entering a viable cell. In the presence of oxygen, it will be re-oxidized and diffuse from the cell. If the cell is hypoxic, however, the radiotracer will be trapped. This tracer was first evaluated in V-79 cells using low O2 levels in the incubation to mimic hypoxia [36] and followed up with cancer models in animals [37]. Glioma patients have undergone [18F]FMISO PET imaging, though with limited success [38]. Chakhoyan et al. were able to build ptO2 maps from [18F]FMISO PET images and compare them to perfusion weighted imaging (MRI) and 1H-MR mono-voxel magnetic resonance spectroscopy (MRS) [39]. The correspondence with MRI and MRS imaging confirms the direct relationship between [ 18F]FMISO and oxygen levels in tissue. However, Valk et al. observed retention of [18F]FMISO in a GBM subject and anaplastic astrocytoma subject, but no retention of [18F]FMISO in another GBM subject [38]. [18F]FMISO PET imaging has a limited range in sensitivity between normoxic and hypoxic tissue [40]. Additionally, BBB penetration is low for [18F]FMISO which does not make it attractive for glioma imaging.

There are multiple hypoxia radiotracers developed around the same time as [18F]FMISO including [ 18F]FET-NIM (**8**), [18F]EF5 (**9**), and [18F]HX4 (**10**) (see review [35]). All of these ligands use the nitroimidazole moiety to sense oxygen level in vivo, and the differences in the chemical structures are primarily focused on the linker groups and location of fluorine-18 label (Figure 2). [18F]EF5 exhibits greater cell membrane permeability, slower clearance, and improved tumor uptake [41–44]. Its structurally very similar to [18F]FMISO and [18F]FETNIM, but incorporates a pentafluoro ethyl group via an amide linker. Labeling at this position requires electrophilic fluorination in the radiosynthesis, limiting regular production of this radiotracer. [18F]FETNIM was developed concurrently with [ 18F]FMISO, and is the most structurally similar with the addition of a hydroxyl group alpha to the nitroimidazole ring [45–47]. [18F]HX4 still contains the nitroimidazole, though added a triazole linker between this and the fluorine-18 label [48–51]. When compared to [18F]FMISO, [18F]HX4 provides the higher image contrast 4 h post-injection, but has high variability [52,53]. A second generation version of [18F]FMISO, [18F]DiFA (**12**) is slightly less lipophilic. This radiotracer aims to have a faster clearance rate, and thus improved signal-to-noise ratio over [18F]FMISO [54].

**Figure 2.** Hypoxia-sensing tracers.

An alternative to [18F]FMISO, and the other organic oxygen sensing compounds, is [64Cu]ATSM (**12**), which uses a metal to sense oxidation changes (Cu(II) to Cu(I)) [55]. In comparison to [18F]FMISO, it has faster tracer kinetics and can reveal 'hypoxic' tissues in 15 min post injection. [64Cu]ATSM uptake is influenced not only by hypoxia, but also cellular concentrations of reducing species such as NADH [56]. Using [64Cu]ATSM imaging in glioma patients, SUV was found to be an independent predictor of both progression free survival and overall survival in this study [57]. In the GBM subgroup analysis, however, max SUV only showed significant prediction of progression free survival [57], using copper-64 allowed for synergy between PET and MRI, in this proof of concept using a closely related [ 64Cu]ATSM compound ([64Cu]Cu(L2); **13**), which has only been evaluated so far in vitro [58,59]. Though, it is likely to have similar BBB penetration issues as [64Cu]ATSM.

Hypoxia imaging techniques are important for the treatment planning of solid tumors to ensure effective regression. While current therapies do not target the hypoxia machinery in cells, this biochemical process remains a critical consideration for radiotherapy. Many tumor types, including gliomas, are described as being hypoxic and utilize PET or MR imaging methods for treatment planning [60]. This method also serves to monitor treatment after targeted radiotherapies.

#### *2.3. Inflammation*

An important part of the immune response is inflammation and, like peripheral cancers, brain cancer cells will trigger this response. A histological analysis of lesions from 1265 patients with glioblastoma multiforme identified the presence of lymphocytes and reactive astrocytes [61]. Inflammation has been explored as a therapeutic target for such cases. Nonsteroidal anti-inflammatory drugs (NSAIDs) have been evaluated to suppress the growth of tumor cells in vitro [62]. While NSAIDs are an inappropriate therapy for GBM because of the lack of targeting, inflammation is a characteristic of GBM pathology and could be useful for imaging.

Translocator protein (18 kDa), TSPO, is an integral mitochondrial membrane protein responsible for cholesterol transport and responds to cell stress. Broadly, TSPO is treated as a biomarker sensitive to pro-inflammatory stimuli, and small molecule inhibitors of TSPO have been utilized in PET imaging [63]. [ 11C]PK11195 (**14,** Figure 3) was developed for general use of inflammation imaging, and even in glioma patients [64]. Immunohistochemical experiments with patient tissue further confirmed increase TSPO mRNA and protein levels [65]. Furthermore TSPO PET imaging was correlated to outcome [66]. In a larger prospective trial which included tumor biopsy, imaging and histology were found to correlate, and *BP*ND in high-grade gliomas was significantly higher than in low-grade astrocytomas and low-grade oligodendrogliomas [67]. Although second and third generation TSPO radioligands with higher specific binding have since been developed, TSPO imaging in GBM has major limitations including tumor heterogeneity and inability to distinguish signal caused by radiation therapy from signal due to the tumor microenvironment.

#### **>&@3.**

**Figure 3.** TSPO ligand.

#### **3. New Biomarkers for GBM PET Imaging**

New strategies for cancer therapeutics target proliferation (sigma 2), immunity (sigma 1, PD-L1), and genetic modification (PARP, IDH). PET imaging agents have developed in tandem with advancement of these therapies. These therapies were initially designed to treat more prevalent, peripheral tumors. The nature of heterogeneity in brain tumors inspires the use of biomarker specific imaging agents, as opposed to the more general ligands for proliferation, hypoxia, and inflammation. Applications to brain cancers comes with the significant considerations of blood–brain barrier penetration and effectiveness. The PET imaging agents discussed in the following section are not yet utilized in human brain imaging but represent promising candidates for the new therapeutic and/or biomarker strategies.

#### *3.1. Sigma 1*

A key characteristic of GBMs is their invasiveness, which leads to the very low survival rate. Sigma 1 and sigma 2 receptors are expressed in the human tumor cell lines: C6 glioma, NIE-115 neuroblastoma, and NG108-15 neuroblastoma- glioma hybrid [68]. While sigma 1 is associated with

invasiveness, sigma 2 receptors have been associated with proliferation. Both receptors are interesting PET imaging targets for glioblastoma.

Sigma 1 receptor (S1R) was identified as the site of action of the antipsychotic haloperidol [69]. Its role in the CNS has been investigated for neurodegenerative disorders [70] in addition to GBM. The first PET ligand for S1R was [11C]1-(3,4-dimethoxyphenethyl)-4-(3-phenylpropyl)piperazine ([11C]SA4503: **15,** Figure 4) [71]. While it showed nanomolar affinity to S1R, it unfortunately has high affinities for other receptors, ion channels, and second messenger systems [72]. Both S1R and S2R have structural similarity to opiate receptors, which necessitates specificity and selectivity when designing ligands. Additionally, in the development of fluorine-18 sigma-1 receptor ligands, very high binding affinity has corresponded to very slow clearance rate. For example, [18F]1-3-fluoropropyl-4-((4-cyanophenoxy)-methyl)piperidine ([18F]FPS; **16**) [73] did not reach pseudo-equilibrium by 4 h in human [74]. Another ligand, [ 18F]6-(3-fluoropropyl)-3-(2-(azepan-1-yl)ethyl)benzo[d]thiazol-2(3H)-one ([18F]FTC-146; **17**) was also found to be irreversible and not suitable for neuroimaging [75]. Recently, [18F](S)-Fluspidine (**18**) [76] was developed and evaluated in human [77]. The pharmacokinetics are improved, and with xenograft mouse models of glioblastoma have visible increases in radioligand binding.

**Figure 4.** Sigma 1 receptor radioligands.

#### *3.2. Sigma 2*

The amino acid and nucleoside base PET imaging approach is a classic, straightforward way to assess functional proliferation. The sigma-2 receptor (S2R; TMEM97) has recently been implicated in cancer biology; S2R levels increased 5-fold in proliferating tumor cells compared to quiescent tumor cells [78]. This allows for radioligand design for a specific protein more akin to drug design. For an in-depth history of S2R ligand development, see references [79,80].

An early imaging effort for S2R was using 4-[125I]BP (**19**, Figure 5), a small molecule with high affinity for both sigma 1 and 2 receptors [81]. Further efforts by these authors led to [125I]PIMBA (**20**), though this radioligand suffered from high background binding [82]. [18F]RHM-4 (**21**) was developed and demonstrated S2R overexpression in pancreatic cancer, though BBB penetration was not investigated [83]. The radioligand [18F]ISO (**22**) was also developed, structurally similar to **21** except lacking the aryl methoxy group [84]. A positive correlation is observed between **22** binding and tumor Ki67 expression [85]. In treatment monitoring of CDK4/6 inhibition plus endocrine therapy in breast cancer xenograft animals, **22** was found to assess more delayed changes related to cell cycle arrest compared to [18F]FLT [86]. However, **22** is not taken up into the brain, based on organ residence studies in rodent [85]. This has inspired alternative ligands with BBB penetration as the design goal.

**Figure 5.** Sigma 2 receptor radioligands.

A scaffold incorporating pthalimides (**23** and **24**) showed elevated brain uptake and specific binding (displaceable with cold ligand and haloperidol), though the tumor to background ratio was low [87]. Abate et al. described **25**, which demonstrated good in vitro binding and specificity; however, as a P-glycoprotein (PGP) substrate it is not suitable for brain imaging [88]. The same group continued with a carbon-11 effort (**26**), based on the inhibitor PB28 and very similar to **19**; however, brain uptake was low and the compound did not display high enough specific binding [89]. Wang et al. reported two fluorine-18 inhibitors, **27** and **28**, with high brain uptake in mice in 2017, though it may be a PGP substrate as well [90]. These efforts have demonstrated some challenges in the radioligand design of sigma 2 receptor inhibitors for brain imaging, i.e., PGP efflux and limited specific binding.

#### *3.3. PD-L1*

Avoiding destruction by immune cells is a powerful strategy utilized by cancer cells. Until recently, imaging research in this space has focused on antibody-based strategies, which is challenging due to limited brain penetration. The molecules used in PET imaging for Programmed death ligand 1 (PD-L1) are large molecule therapeutics, including antibodies, antibody fragments, and peptides. An early imaging effort developed at Johns Hopkins University adapted the therapeutic antibody for PET imaging: [64Cu]Azetozolizumab [91]. The same group later developed a peptide, [64Cu]WL12, although brain was not listed in the biodistribution study [92]. Another protein effort from the Gambhir lab at Stanford, [64Cu]NOTA-HACA-PD1 and a gallium 68 version, determined no brain uptake definitively [93]. Merck developed an affibody ZPD-L1\_1 that was fluorine-18 labeled; however, brain

penetration has not been demonstrated, though an affibody is more likely to be BBB penetrant than the preceding antibodies [94].

Although small molecule inhibitors have been in development, to date no small molecule radioligands have been described. Bristol Meyers Squib has the first small molecule inhibitor reported, BMS-202 (**29**, Figure 6) [95] and has characterized the binding in a crystal structure [96]. The mechanism of action is thought to be selectively induced dimerization of PD-L1, which inhibits binding to PD-1 (see reviews [97–99])

**Figure 6.** Small molecule PD-L1 inhibitor.

#### *3.4. PARP*

ADP-ribose polymerase (PARP) is the enzyme responsible for attaching linear or branched polymers of ADP onto broken DNA and other biomolecules. PARP-1 recognizes single and double-strand breaks, crossovers, cruciform, and supercoils. Additionally, PARP-1 maintains the stability of replication forks, and the basal activity is very low [100]. In cancer cells containing a BRCA2 deficiency, inhibiting PARP causes synthetic lethality [101]. This treatment strategy has been used to develop multiple small molecule therapeutics [102].

The first in-class inhibitor olaparib has been adapted for use in PET imaging: [18F]BO (biorthogonal olaparib; **30**, Figure 7) [103]. [18F]BO demonstrated ubiquitous distribution in cancer cells and localization within the nucleus in cancer cells [104]. Structurally similar to olaparib, the inhibitor radioligand [18F]PARPi (**31**), and corresponding fluorescent version (**32**), has been used in GBM cell lines [104] and in rodent [105]. Although this shows specific binding in peripheral tumors, there was no significant blocking with brain uptake at 2 h biodistribution. Brain penetration was very low initially. In comparison to [11C]choline and [18F]FLT, PARPi demonstrated a lower mean uptake in tumor than the other two PET ligands; however, the lower background uptake enabled PARPi to delineate brain tumors in rodent models with more clear contrast [106]. The Gouverneur group recently radiolabeled olaparib itself with fluorine-18, though limited imaging studies have been done with this tracer [107]. As expected, [18F]olaparib (**33**) was taken up selectively in PARP-1 expressing cells and in mouse tumors. Additionally, radioligand uptake was increased by 70% after tumor irradiation, indicated a great potential for monitoring radiation damage. However, brain was not included in biodistribution calculations and there does not appear to be brain uptake in the dynamic PET images. Unlike the preceding PARP radioligands, the inhibitor [18F]fluorthanatrace ([18F]FTT; **34**) is not based on olaparib, but rather rucaparib. This benzimidazole carboxamide derivative is highly potent with an IC50 of 6.3 nM [108]. It has been evaluated in humans, though with low brain penetration [109]. There remains a need for BBB penetrating small molecule inhibitors of PARP in order to be useful specifically in glioblastoma imaging.

**Figure 7.** PARP ligands.

#### *3.5. Isocitrate Dehydrogenase (IDH)*

Through genome wide association studies (GWASs), the common mutation R132H, located on the isocitrate dehydrogenase (IDH) 1 gene, was found in more than 70% of grade II and III astrocytomas, oligodendrogliomas, and glioblastomas that developed from these lower-grade lesions [110]. However, only 10% of glioblastomas are IDH mutant [1], and this is weakly associated with tumor aggressiveness [3]. IDH- mutant GBM has a significantly longer survival rate compared to IDH wild type, 31 months compared to 15 months with standard treatment.

The observed IDH mutation essentially eliminates all enzymatic activity [111]. In normal cells, IDH1 converts isocitrate to α-ketoglutarate and form NADPH, which maintains a pool of reduced glutathione and peroxiredoxin. When isocitrate cannot be converted to α-ketoglutarate, it instead is converted to R(-)-2-hydroxyglutarate [112]. 2-Hyrdoxyglutarate and α-ketoglutarate are cofactors for many enzymes and their availability influences DNA methylation status. 2-HG inhibits 5mC hydroxylase (TET2) and lysine demethylases (KDM) leading to demethylation of DNA and histone, respectively. This changes gene expression and thus tumorigenesis. R-2HG stimulates EgIN1, which promotes HIF1a degradation by hydroxylation [113]. Therapeutic inhibitors are being developed for IDH1 to influence this pathway and similarly slow tumorigenesis and promote survival.

Ivosidenib (AG-120) is the first in-class IDH1 reversible inhibitor [114]. In rats, some brain penetration has been demonstrated, 4.1% after 50 mg/kg dose, which could indicate some effectiveness in glioblastoma. The phase I trial in low grade glioma subjects is ongoing (NCT03343197). Another inhibitor (IDH305) for mutant IDH1 is in development by Novartis, with strong data for binding in brain homogenate [115]. In a phase I clinical trial, safety was evaluated in glioma, AML/MDS, and other/ non-CNS solid tumors with IDH mutation [116]. However, phase II trials of IDH305 in glioma has been withdrawn, possibly due to liver toxicity (NCT02977689 [117]).

Before small molecule inhibitors of IDH were available, PET radioligands were evaluated for their relationship to IDH mutation status. While [18F]FDOPA does not correlate with IDH mutation status [118], it appears that [18F]FET imaging does significantly associate with IDH mutation status [119–121], although there does not appear to be a biochemical relationship between [18F]FET uptake and IDH mutation status. With the disclosure of small molecule IDH inhibitors came the preliminary development of IDH-selective PET radioligands. Chitneni et al. utilized an iodine-131 and fluorine-18 version (**35**, Figure 8) of AGI-5198; however, these compounds lacked selectivity for mutant IDH over wild type [122]. In a follow up from the same group, [18F]triazinediamine (**36**) analogues were radiosynthesized, based on Enasidenib (AG-221) [123], The Kd (**36**) was calculated to be 40 nM with a Bmax of 4426 gmol/mg in a mutant anaplastic astrocytoma cell line, which is promising. However, biodistribution studies showed bone uptake from radiolytic defluorination, so further design is required [123].

**Figure 8.** Isocitrate dehydrogenase (IDH) radioligands.

#### **4. General Imaging Considerations**

Brain PET imaging for any disease state is challenging. The blood–brain barrier (BBB) poses a major obstacle for any successful radiotracer targeting glioblastoma. Even if the radiotracer enters the brain, many compounds exhibit slow brain entry (i.e., small *K*<sup>1</sup> values). Since glioblastoma lesions often compromise BBB, increased radioactivity concentrations at the lesion site may reflect increased nonspecific radiotracer instead of, or in addition to, increased target signal. For example, a combined [ 18F]FMISO PET and MRI study noted high uptake of [18F]FMISO in areas of BBB disruption as well as in necrotic tissue [124]. This raises a challenge in quantification at a suspected glioblastoma lesion. Kinetic modeling approaches incorporating dynamic data, in some cases, may help distinguish nonspecific signal from specific signal in some cases. Although these scanning protocols can require longer scanning time and possibly arterial blood sampling, the potential for improved quantification of specific radiotracer uptake offers important benefits to evaluating diagnosis, staging, or treatment efficacy that must be considered when using PET to image GBM.

Reference region approaches offer alternatives for quantitative analyses that may shorten the length of scanning time and do not require arterial blood sampling. While such approaches have major limitations in cases of ubiquitously expressed targets, such as TSPO, in the case of brain tumor imaging the reference region may be drawn as a larger region that is removed from the lesion. For example, in a case of [18F]FET imaging a fixed-size reference ROI was placed in hemispheres contralateral to tumorous tissue, which yielded fully image-derived measures that correlated with disease-free survival [125]. Image-derived input function allow for alternative non-invasive modeling approaches. For example, [18F]FMISO has no reference region, but image-derived tissue-to-blood ratios provide reasonable proxies for measured parent radioactivity in venous blood [126]. Such approaches can maintain quantitative accuracy while reducing logistical complexities introduced by full dynamic scans with blood sampling.

A final major obstacle for PET radiotracers imaging GBM is high nonspecific binding. For example, [ 18F]FMISO, [18F]FLT, and [64Cu]ATSM exhibit elevated nonspecific binding, which limits their usefulness. Such radiotracers exhibit low signal to noise ratio (SNR), which makes it more difficult to differentiate areas of low uptake from noise. In contrast, radiotracers with low nonspecific binding may be amiable to reference region analysis if non-tumor brain regions exhibit negligible specific binding. Off-target binding poses another related challenge. For example, sigma 1 and sigma 2 receptors have significant structural similarities to opiate receptors. These challenges highlight the need for blocking studies with candidate radiotracers to confirm suitable sensitivity and specificity without high nonspecific binding.

#### **5. Conclusions**

PET imaging provides important functional information about GBM tumors and surrounding tissue environment. Markers of proliferation, hypoxia, and inflammation have been used to image the lesions of GBM patients. Exciting new frontiers for PET imaging targets for GBM include PD-L1 for immune status, PARP-1 for DNA damage, sigma 2 receptors as alternative markers of proliferation, and isocitrate dehydrogenase for tumorigenesis activity. These PET imaging targets have the potential to enhance diagnosis, staging, and treatment approaches for GBM. As GBM PET imaging techniques advance, it is critical to consider blood brain barrier penetration and nonspecific binding in the evaluation of new radioligands.

**Author Contributions:** L.R.D. performed literature search and drafted the initial manuscript. Z.C. and A.T.H. revised and edited subsequent manuscript iterations. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by NIH grants K01AA024788, R21EB027872, R01AG058773. Z.C. is an Archer Foundation Research Scientist.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Molecules* Editorial Office E-mail: molecules@mdpi.com www.mdpi.com/journal/molecules

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18