Development of a PET radioligand for α2δ-1 subunit of calcium channels for imaging neuropathic pain

Abstract

Neuropathic pain affects 7-10% of the adult population. Being able to accurately monitor biological changes underlying neuropathic pain will improve our understanding of neuropathic pain mechanisms and facilitate the development of novel therapeutics. Positron emission tomography (PET) is a noninvasive molecular imaging technique that can provide quantitative information of biochemical changes at the whole-body level by using radiolabeled ligands. One important biological change underlying the development of neuropathic pain is the overexpression of α2δ-1 subunit of voltage-dependent calcium channels (the target of gabapentin). Thus, we hypothesized that a radiolabeled form of gabapentin may allow imaging changes in α2δ-1 for monitoring the underlying pathophysiology of neuropathic pain. Here, we report the development of two ¹⁸F-labeled derivatives of gabapentin (trans-4-[¹⁸F]fluorogabapentin and cis-4-[¹⁸F]fluorogabapentin) and their evaluation in healthy rats and a rat model of neuropathic pain (spinal nerve ligation model). Both isomers were found to selectively bind to the α2δ-1 receptor with trans-4-[¹⁸F]fluorogabapentin having a higher affinity. Both tracers displayed around 1.5- to 2-fold increased uptake in injured nerves over the contralateral uninjured nerves when measured by gamma counting ex vivo. Although the small size of the nerves and the signal from surrounding muscle prevented visualizing these changes using PET, this work demonstrates that fluorinated derivatives of gabapentin retain binding to α2δ-1 and that their radiolabeled forms can be used to detect pathological changes in vitro and ex vivo. Furthermore, this work confirms that α2δ-1 is a promising target for imaging specific features of neuropathic pain.

1. Introduction

Defined by the International Association for the Study of Pain (IASP), neuropathic pain is the “pain caused by a lesion or disease of the somatosensory nervous system” [1]. An epidemiological study suggests that the prevalence rate for neuropathic pain is between 6.9% and 10% [2]. The clinical assessment of pain is mainly based on self-report [3] whereas the preclinical assessment relies on measuring physiological/behavioral response to evoking stimuli [4, 5]. Pain imaging using electroencephalography (EEG), magnetic resonance imaging (MRI), or positron emission tomography (PET) can provide additional information of structure, function, or function underlying the pain condition, which can help us better understand the mechanisms and biochemical processes contributing to the pathological pain [6–10] and help us develop nonopioid analgesics [11, 12]. Compared with EEG and MRI, PET can be used to visualize biochemical events in vivo at a whole-body dimension and to quantify the density and occupancy of the neuroreceptors or therapeutic targets with high sensitivity and moderate spatial resolution [13]. PET tracers such as [¹⁸F]FDG, [¹¹C]diprenorphine/[¹⁸F]FDPN, 6-[¹⁸F]fluoro-L-DOPA, [¹¹C]ABP688, and [¹⁸F]FTC-146 have been used to image glucose uptake [14–16], opioid receptor density [17–19], dopaminergic activity [20], glutamate receptor 5 (mGluR5) binding level [21], and the expression of sigma 1 receptor [22–24] under different pain conditions. However, the development of neuropathic pain involves other pathways and receptors [25] which have not been sufficiently studied by PET imaging. Therefore, developing novel tracers for other targets specific to neuropathic pain is pressingly needed.

The α2δ subunit is one of the auxiliary components of voltage-gated calcium channels (VGCCs). α2δ subunits have four gene types, which encode subunits α2δ-1, α2δ-2, α2δ-3, and α2δ-4. α2δ-1 subunit is mainly expressed in muscles, central nervous system (CNS), peripheral nervous system (PNS), and endocrine tissues [26]. Multiple studies have shown that the expression of α2δ-1 is highly increased (3-17 fold) in the nerves, dorsal root ganglia and spinal cords of different animal models of neuropathic pain [27–37]. In addition, transgenic mice overexpressing α2δ-1 display a phenotype of hyperalgesia [38] and α2δ-1 knock-out mice show reduced sensitivity to mechanical and cold stimuli [39]. Remarkably, the increased expression of α2δ-1 in the spinal dorsal horn of neuropathic pain animals can be reversed by treatment with gabapentinoids [30, 40, 41]. Taken together, these results demonstrate that α2δ-1 is a robust biomarker of neuronal injury associated with neuropathic pain. Despite the potential value of imaging α2δ-1 receptor by PET, this has not been achieved, mainly due to the lack of a suitable PET tracer. In 2019, Yang et al. communicated at the SNMMI annual meeting the development of a copper-64 labeled antibody for the α2δ-1 receptor [42]. This antibody-based PET tracer showed promise for cancer imaging, however its ability to detect changes in nerves or spinal cord due to neuropathic pain has not been reported.

Gabapentinoids are ligands of α2δ receptors and the first-line treatment of neuropathic pain recommended by the International Association for the Study of Pain (IASP) Neuropathic Pain Special Interests Group (NeuPSIG) and European Federation of the Neurological Societies (EFNS) [25]. FDA approved gabapentinoids include gabapentin (commercial name Neurontin) and pregabalin (commercial name Lyrica) (Scheme 1). Given the efficacy of these compounds in neuropathic pain and their selectivity towards α2δ receptors, we hypothesized that radiolabeled gabapentinoids may serve as PET tracers for neuropathic pain imaging. In particular, gabapentin appears to have suitable properties including high affinity towards α2δ-1 receptor, high metabolic stability, and is amenable to chemical modifications [43, 44].

Scheme 1.

Chemical structures of gabapentin, pregabalin, and 4-fluorogabapentin.

Furthermore, fluorine-18 labeled tracers are generally preferred over carbon-11 tracers as their longer half-life (110 min vs. 20 min) allows for remote production and regional distribution. Therefore, we set out to develop novel fluorinated derivatives of gabapentin and their ¹⁸F-labeled versions. In this paper, we describe the synthesis, and the evaluation of two novel radiotracers in healthy rats and rats post spinal nerve ligation (SNL), a model of peripheral neuropathic pain.

2. Results and discussion

2.1. Validation of gabapentin as a basis for a neuropathic pain tracer

As described above, the expression of α2δ-1 increases in animal models of peripheral neuropathic pain (Fig. 1A) [27–37]. Before investing effort in generating new ¹⁸F-labeled ligands and testing them in vivo, we decided to test whether the increased expression in α2δ-1 in the spinal cord of mice with neuropathic pain would result in a measurable increase in gabapentin binding. To do this, we incubated spinal cord sections obtained from spared nerve injury (SNI) mice, a well-established model of neuropathic pain in which two of the three distal branches at the foot level of the sciatic nerve are transected and ligated [45], with tritium labeled gabapentin ([³H]GBP). Incubation of longitudinal spinal cord sections showed a 245% increase in [³H]GBP binding in the spinal cord of SNI animals compared to sham operated mice (Fig. 1B). Autoradiography of axial spinal cord sections showed a binding pattern consistent with the previously reported expression of α2δ-1 [46] and a 160% higher binding in SNI animals than controls (Fig. 1C). Furthermore, closer inspection of the signal in SNI sections showed 38% higher binding in the dorsal horn ipsilateral to the injured nerve than in the dorsal horn of the contralateral side (Fig. 1D).

Figure 1.

Increased binding of [³H]gabapentin in the spinal cord of neuropathic pain mice. A. Diagram illustrating the neuropathic pain models used in this investigation: spared nerve injury (SNI) model is produced by lesioning the common peroneal (CPN) and tibial nerve (TN) branches of the sciatic nerve while sparing the sural nerve (SN); spinal nerve ligation (SNL) model is produced by ligating the L5 and L6 spinal nerves. These injuries result in increased α2δ-1 expression in the dorsal root ganglia (DRGs), sciatic nerve and spinal cord. B. Representative in vitro [³H]GBP autoradiography of longitudinal spinal cord sections of spared nerve injury (SNI) mice and sham controls. SNI animals showed a 2.45-fold increase in [³H]GBP binding in whole spinal cord sections. C. Representative in vitro [³H]GBP autoradiography of axial spinal cord sections of spared nerve injury (SNI) mice and sham controls. SNI animals showed a 1.6-fold increase in [³H]GBP binding in whole spinal cord sections. D. Representative zoom in in vitro [³H]GBP autoradiography images of axial spinal cord sections of spared nerve injury (SNI) mice and sham controls. SNI animals showed a 1.4-fold higher [³H]GBP binding in the dorsal horn ipsilateral to the injury than the contralateral side.

2.2. Design and (radio)synthesis of 4-fluorogabapentin and 4-[¹⁸F]fluorogabapentin

Encouraged by the ability of [³H]GBP to detect changes in a neuropathic pain model, we decided to pursue fluorinated derivatives of gabapentin. Fluorinated derivatives of gabapentin have not been previously described. Previous structure-activity relationship (SAR) studies of gabapentin alkyl derivatives have indicated that the trans-4-substituted, (1S,3R)-, and (1R,3S)-3-substituted gabapentin (substitution = methyl or ethyl) show similar binding affinities to the α2δ-1 compared to the parent drug gabapentin [47]. Based on these SAR, we hypothesized that addition of a fluorine atom to the 4-position would retain binding to the target. The methods to synthesize 4-fluorogabapentin both with nonradioactive fluorine-19 and radioactive fluorine-18 are discussed here.

As shown in Scheme 2, the synthesis of nonradioactive 4-fluorogabapentin starts from the commercially available precursor 2-azaspiro[4.5]decane-3,8-dione (1). First, the amide -NH is protected by tert-butyloxycarbonyl (Boc) by the reaction of 1 with di-tert-butyl dicarbonate (Boc₂O). The subsequent reduction of the Boc-protected intermediate 2 with excess amount of sodium borohydride (NaBH₄) in methanol (MeOH) yielded the trans- and cis- mixture of compound 3 in 97% chemical yield. The isolation of trans- (3b) and cis- (3a) isomers was achieved by reverse-phase semi-preparative HPLC. Deoxyfluorination of 3a and 3b using DAST (diethylaminosulfur trifluoride) or AlkylFluor yielded the fluorinated Boc-protected gabapentin lactam 4a and 4b, respectively. AlkylFluor showed better deoxyfluorination efficiency (40-50%) than DAST (10%), however, it also resulted in more side products. Finally, the hydrolysis and deprotection of 4a and 4b in one pot yielded the final products trans-4-fluorogabapentin (5a, tGBP4F) and cis-4-fluorogabapentin (5b, cGBP4F) in 70-80 % yield.

Scheme 2.

Synthesis of trans-4-fluorogabapentin (5a, tGBP4F) and cis-4-fluorogabapentin (5b, cGBP4F).

In order to prepare the ¹⁸F-labeled products, cis/trans mixture 3 or isomerically pure 3a/3b were reacted with MsCl (methanesulfonyl chloride) in dichloromethane in the presence of NEt₃ to give the mesylated compounds 6 and 6a/6b (Scheme 3). Stereochemically pure 6a and 6b can be obtained both through direct mesylation of the stereochemically pure precursors (3a/3b) or by the semi-prep reverse-phase HPLC purification of compound 6. The stereochemical assignment of 6a and 6b was determined based on the large vicinal H-H coupling [48] of 4-position proton and 3/5-position proton of 6a [H₄ δ = 4.77 ppm (tt, J₁ = 4 Hz, J₂ = 8 Hz, 1 H)] and small vicinal H-H coupling of 4-position proton of 6b [H₄ δ = 4.80 ppm (tt, J₁ = 5 Hz, J₂ = 5 Hz, 1 H)] as well as the NOESY H-H coupling of lactam ring proton with axial proton of cyclohexyl ring (NMR spectra section, Supporting Information). Both assignment methods provided consistent and unequivocal results. The stereochemical assignments of 3a/3b, 4a/4b, and 5a/5b were based on the assignment of 6a/6b after confirming experimentally that these transformations were stereoselective.

Scheme 3.

Synthesis of trans-4-[¹⁸F]fluorogabapentin ([¹⁸F]tGBP4F, 8a) and cis-4-[¹⁸F]fluorogabapentin ([¹⁸F]cGBP4F, 8b) (compounds 7, 7a, and 7b were not isolated).

Fluorine-18 labeled gabapentin analogues were synthesized by nucleophilic substitution of [¹⁸F]F⁻ of the mesylate precursors 6 or 6a/6b. As shown in Scheme 3, labeling of precursor 6 led to the isomer mixture [¹⁸F]fluorogabapentin (8, [¹⁸F]GBP4F) which was further purified to obtain the stereochemically pure trans-4-[¹⁸F]fluorogabapentin (8a, [¹⁸F]tGBP4F) and cis-4-[¹⁸F]fluorogabapentin (8b, [¹⁸F]cGBP4F). On the other hand, labeling of the stereochemically pure precursors 6a and 6b yielded stereochemically pure trans-4-[¹⁸F]fluorogabapentin (8a, [¹⁸F]tGBP4F) and cis-4-[¹⁸F]fluorogabapentin (8b, [¹⁸F]cGBP4F) directly (Scheme 3).

Labeling of precursor 6 with 18.5-55.5 GBq (0.5–1.5 Ci) of [¹⁸F]KF generated [¹⁸F]GBP4F (8) in 5.8 ± 1.8 % (n = 9) decay corrected radiochemical yield and >99% of radiochemical purity in ~120 min of synthesis and purification time. The ratio of [¹⁸F]tGBP4F to [¹⁸F]cGBP4F was ca. 2:1. Labeling of precursor 6a with 11.1-37 GBq (0.3–1.0 Ci) of [¹⁸F]KF generated [¹⁸F]tGBP4F (8a) in 4.3 ± 2.2 % (n = 13) decay corrected radiochemical yield and >99% of radiochemical purity. Labeling of precursor 6b generated [¹⁸F]cGBP4F (8b) in 1.4 ± 1.2 % (n = 3) decay corrected radiochemical yield and >99% of radiochemical purity. Due to the low UV-Vis absorbance of the 4-fluorogabapentin, the identity of the radioactive compounds was confirmed by coinjection with reference standards on analytical radioHPLC and monitoring by evaporative light-scattering (ELS) detection (Fig. S1, Supporting Information).

2.3. In vitro evaluations

In order to assess the target binding of the newly synthesized α2δ-1 ligands tGBP4F and cGBP4F, competitive radioligand binding assays to rat spinal cord sections were carried out and analyzed via quantitative in vitro autoradiography.

As shown in Fig. 2A, the autoradiography of free [¹⁸F]tGBP4F and [¹⁸F]cGBP4F showed an accumulation of radioactivity in the dorsal horn of the spinal cord sections, consistent with immunohistochemistry (Fig. S2, Supporting Information). This was also consistent with previously reported α2δ-1 immunohistochemistry results [46] and the [³H]gabapentin autoradiography results (Fig. 1). As shown in Fig. 2A, >90% binding of [¹⁸F]tGBP4F and [¹⁸F]cGBP4F could be displaced by 10 μM of tGBP4F/cGBP4F or 10 μM of GBP, which demonstrates high specific binding.

Figure 2. Binding of tGBP4F and cGBP4F to α2δ-1 receptors.

A: Comparison of autoradiography of [¹⁸F]tGBP4F (left: baseline; right: blocking w/ 10 μM tGBP4F) and [¹⁸F]cGBP4F (left: baseline; right: blocking w/ 10 μM cGBP4F) in healthy rat spinal cord slices. Areas of high 4-[¹⁸F]fluorogabapentin binding correlate with published IHC results of high α2δ-1 expression area.[46] B-D: Quantitative autoradiography of [¹⁸F]tGBP4F with increasing concentrations of tGBP4F (B), GBP (C), and cGBP4F (D) in healthy rat spinal cord slices. E: Competition binding curve for the inhibition of the [¹⁸F]tGBP4F binding radioligand binding by increasing concentrations of tGBP4F, GBP, and cGBP4F.

To assess the relative potency of the gabapentin derivatives, the half maximal inhibitory concentration (IC₅₀) of tGBP4F, cGBP4F and GBP was measured by competitive radioligand binding against [¹⁸F]tGBP4F. Rat spinal cord sections were incubated with [¹⁸F]tGBP4F and increasing concentrations of non-radioactive competitors tGBP4F, GBP, and cGBP4F. As shown in Fig. 2B–2D, the [¹⁸F]tGBP4F binding was gradually inhibited by the non-radioactive competitors. The IC₅₀ values of each competitors were calculated accordingly (Fig. 2E, IC₅₀: tGBP4F = 28 ± 15 nM, n = 4; GBP = 38 ± 16 nM, n = 4; cGBP4F = 2415 ± 535 nM, n = 4). These results indicated that trans-isomer tGBP4F shows higher binding affinity than parent drug gabapentin and that the cis-isomer cGBP4F shows significantly lower binding affinity than the trans-isomer. This observation is consistent with previously reported structure-activity relationship studies of alkyl derivatives of gabapentin, in which trans-4-substituted gabapentin (substitution = methyl or ethyl) showed slightly lower binding affinity than GBP whereas the cis-4-substituted gabapentin (substitution = methyl) showed significantly lower affinity [47].

2.4. In vivo evaluation in healthy rats

Being able to produce the ¹⁸F-labeled versions and encouraged by the selective and potent binding of [¹⁸F]tGBP4F and [¹⁸F]cGBP4F, we set out to evaluate these novel radioligands in live rats. A 60 minute dynamic whole body PET scans followed by a 15 minute CT scan was carried out on tracer-injected rats using a microPET/CT scanner. As shown in Fig. 3A and Sup. Fig. S3, both radiotracers showed wide distribution throughout the body with highest signal in bladder and kidneys. In the brain, [¹⁸F]tGBP4F showed a standard uptake value (SUV) of 0.7 at 10 min which decreased to 0.4 by 50 min. In contrast, [¹⁸F]cGBP4F brain SUV was 0.7 at 10 min and increased to 0.9 by 50 min. Higher SUV values and similar trend over time was observed in muscle tissue (Fig. 3B). In the kidneys, the signal from [¹⁸F]tGBP4F and [¹⁸F]cGBP4F decreased from SUV = 17 at 10 min to SUV = 8 at 50 min and from 14 to 12, respectively. Consistently, the signal in the bladder for [¹⁸F]tGBP4F and [¹⁸F]cGBP4F increased from 30 to 105 SUV and 11 to 26 SUV within the first 60 min, respectively. Taken together, these results showed moderate brain uptake of both radioligands and faster clearance of [¹⁸F]tGBP4F compared to [¹⁸F]cGBP4F. The brain uptake of these derivatives is comparable to that of gabapentin [49, 50], which is imported in the brain through the LAT1 transporter [51].

Figure 3. [¹⁸F]tGBP4F and [¹⁸F]tGBP4F in rats.

A: Whole body PET imaging of healthy rats (coronal view, summed images at 0–20 min, 20–40 min, and 40–60 min intervals) with [¹⁸F]tGBP4F (left) and [¹⁸F]cGBP4F (right); B: PET imaging based time-activity curves of [¹⁸F]tGBP4F and [¹⁸F]cGBP4F in healthy rats. C: Biodistribution of [¹⁸F]tGBP4F and [¹⁸F]cGBP4F in brain, thoracic spinal cord (tSpC), lumbar spinal cord (lSpC), muscle, and L4-L6 spinal nerves of healthy rats (blocking dose: 30mg/kg of gabapentin, 30 min pre-injection) at 75 min post tracer injection (data obtained by the gamma counting of the dissected tissues, *: p < 0.05).

Due to the relatively small size of the rat spinal cord and spinal nerves, it is very difficult to accurately measure the SUV of these tissues using microPET imaging. To accurately measure the SUV in these tissues, rats were injected with the tracer and euthanized 75 min after injection. Their tissues were immediately dissected and the radioactivity in the tissues was measured by gamma counting. SUVs were calculated accordingly. This was done both under baseline (tracer-only) condition as well as 30 min after IP injection of 30 mg/kg of GBP (blocking) to assess specific binding in vivo. A dose of 30 mg/kg was chosen as this is the lowest dose that has been reported to be effective in alleviating signs of neuropathic pain in rodent models [52]. As shown in Fig 3C and consistent with the previous PET results, [¹⁸F]tGBP4F showed lower SUV than [¹⁸F]cGBP4F across tissues. The signal in the thoracic spinal cord was similar to that of the brain and the signal in the lumbar spinal cord was 1.7-2.4-fold greater. The signal in the spinal nerves was similar to that of the lumbar spinal cord. Finally, the signal in muscle (where α2δ-1 receptors are also highly expressed) was 10-20% greater than the signal in the nerves and lumbar spinal cord. Under blocking conditions, the signals of [¹⁸F]tGBP4F and [¹⁸F]cGBP4F in the muscle showed a −21% (n = 5, p < 0.05) and −16% (n = 5, p < 0.05) decrease, respectively. In the L4-L6 spinal nerves, [¹⁸F]cGBP4F also showed a −21% decrease (n = 10, p < 0.05) but this decrease was not observed with [¹⁸F]tGBP4F. Finally, the signals of both tracers in the brain and spinal cord did not show significant difference between baseline and blocking condition possibly due to the lower brain and spinal cord tracer uptake compared to muscle and peripheral nerves.

2.5. In vivo and ex vivo evaluation in animal model of neuropathic pain (SNL rats)

To check if the tracers could detect increases in α2δ-1 expression under a neuropathic pain condition, both tracers were evaluated in rats after a spinal nerve ligation (SNL). The ligation was performed by unilaterally ligating the L5 and L6 spinal nerves [53]. One-hour dynamic PET scans with [¹⁸F]tGBP4F and [¹⁸F]cGBP4F were carried out using a microPET/CT scanner ~2 weeks after surgery. PET imaging showed similar results to healthy rats and given the low CNS permeability and high background signal from surrounding muscle and the nearby bladder, no measurable changes at the ligation sites could be detected (Fig. S4, Supporting Information). Therefore, biodistribution and ex vivo autoradiography experiments were carried out to monitor tracer binding in SNL rat brains, spinal cords and spinal nerves. The rats were euthanized 75 min after the tracer injection. Selected tissues and the L4-L6 spinal nerves on both the ligated side and unligated side were collected, weighed, and measured in a gamma counter for activity. Brain, thoracic spinal cord (tSpC), lumbar spinal cord (lSpC), and control (unligated) L4-L6 spinal nerves showed similar SUVs as healthy rats (Fig 4A). Compared to control nerves, [¹⁸F]tGBP4F showed around a 2-fold higher uptake in ligated nerves (205 ± 51%, n = 5, Fig. 4C). Similarly, [¹⁸F]cGBP4F showed around a 1.6-fold higher uptake in ligated vs. contralateral control nerves (166 ± 22 %, n = 4, Fig. 4E). Consistent results were obtained for both tracers when the activity was measured using ex vivo autoradiography instead of gamma counting (Fig. 4B and 4D). Importantly, the SUV of ligated nerves was around 11% higher than the surrounding muscle for [¹⁸F]cGBP4F and 20% higher for [¹⁸F]tGBP4F suggesting that in larger animals and humans, where the size of the spinal cord and the spinal nerves is larger than the resolution of PET, these tracers may serve to detect injured nerves.

Figure 4.

A: Biodistribution of [¹⁸F]tGBP4F and [¹⁸F]cGBP4F in select organs/tissues of SNL rats at 75 min post tracer injection (data obtained by the gamma counting of the dissected organs/tissues). B: Representative ex vivo (top) and in vitro (bottom) autoradiography of [¹⁸F]tGBP4F with SNL rat spinal nerves (10-15 days post-surgery) indicates an increase of [¹⁸F]tGBP4F accumulation to ligated nerves (right nerve) compared to non-ligated nerves (left nerve). C: The gamma counting shows 205 ± 51% increase of [¹⁸F]tGBP4F accumulation of SNL nerves (right nerves) compared to healthy nerves (left nerves). D: Representative ex vivo (top) and in vitro (bottom) autoradiography of [¹⁸F]cGBP4F with SNL rat nerves indicates an increase of [¹⁸F]cGBP4F accumulation to ligated nerves (right nerves) compared to non-ligated nerves (left nerves). E: The gamma counting shows 166 ± 22% increase of [¹⁸F]cGBP4F binding to SNL nerves (right nerves) compared to healthy nerves (left nerves).). F: Representative in vitro autoradiography of [¹⁸F]tGBP4F with healthy (left slide) and SNL (right slide) rat spinal cord slices. Healthy rat showed same [¹⁸F]tGBP4F binding in left and right dorsal horn of the spinal cord sections. SNL rat showed higher [¹⁸F]tGBP4F binding in right dorsal horn (same as ligation side) of the spinal cord sections. G: Quantitative autoradiography data shows 148 ± 15% (n = 8) increase of [¹⁸F]tGBP4F accumulation in unilateral side of dorsal horn of SNL rat spinal cord section. H: Representative in vitro autoradiography of [¹⁸F]cGBP4F with healthy (left slide) and SNL (right slide) rat spinal cord slices. Healthy rat showed same [¹⁸F]cGBP4F binding in left and right dorsal horn of the spinal cord sections. SNL rat showed higher [¹⁸F]cGBP4F binding in right dorsal horn (same as ligation side) of the spinal cord sections. I: Quantitative autoradiography data shows 112 ± 5% (n = 8) increase of [¹⁸F]cGBP4F accumulation in unilateral side of dorsal horn of SNL rat spinal cord section.

In vitro autoradiography with both tracers in explanted nerves also showed 1.5-2 fold higher binding in the injured than control nerves (Fig. 4B and 4D). This finding confirms that the increased uptake measured ex vivo is due to higher binding and not due to greater blood flow. Overall, the increase binding of the tracers in injured nerves is consistent with the literature on increased expression of the α2δ-1 receptor in spinal nerve ligation animals [31, 33, 41, 54–58].

In addition, in vitro quantitative autoradiography with both tracers in spinal cord sections of SNL rats showed higher binding in the dorsal horn ipsilateral to the injured nerve (Fig. 4F–4I). These observations are supported by the immunohistochemistry staining (Fig S5, Supporting Information) and consistent with the overexpression of α2δ-1 receptors throughout the nerve as well as the ipsilateral dorsal root ganglia (DRG) and spinal cord of the SNL model as previously reported [29, 46, 55].

3. Conclusion

In conclusion, two novel fluorine derivatives of gabapentin, namely trans-4-fluorogabapentin (tGBP4F) and cis-4-fluorogabapentin (cGBP4F), were synthesized and characterized. Both compounds bind to the α2δ-1 receptor. tGBP4F shows a slightly higher IC₅₀ compared to gabapentin suggesting it also has potential as a therapeutic. Fluorine-18 versions, trans-[¹⁸F]4-fluorogabapentin ([¹⁸F]tGBP4F) and cis-[¹⁸F]4-fluorogabapentin ([¹⁸F]cGBP4F) were also developed for α2δ-1 receptor imaging. Although the chemical structures of the two isomers, [¹⁸F]tGBP4F and [¹⁸F]cGBP4P are highly similar, their performance is significantly different (Table 1). In terms of similarities, both tracers show specific binding to α2δ-1 receptor in vitro as evidenced by the fact that both can be displaced by non-radioactive GBP (Fig 2). In terms of differences, the radiolabeling yield of [¹⁸F]tGBP4F is almost 3-fold higher than [¹⁸F]cGBP4F. In addition, [¹⁸F]tGBP4F shows about 100-fold higher affinity towards α2δ-1 receptor than [¹⁸F]cGBP4F (even higher than gabapentin) (Fig 2). Furthermore, in vivo evaluation shows that [¹⁸F]tGBP4F has a faster clearance than [¹⁸F]cGBP4F, which is generally advantageous for PET tracers (Fig 3). Moreover, both tracers show increased binding in injured nerves with [¹⁸F]tGBP4F being higher than [¹⁸F]cGBP4F when measured by ex vivo gamma counting and ex vivo autoradiography (Fig 4). Due to the small size of rat spinal nerves, and the relatively high signal in surrounding tissues, these changes were not observable by PET. However, we speculate that those changes might be observable in larger animals and humans.

Table 1.

Comparison of [¹⁸F]tGBP4F) and [¹⁸F]cGBP4F.

	Radiosynthesis	α2δ-1 affinity	Pharmacokinetics	CNS permeability
[18F]tGBP4F	high yield	high	slow	moderate-low
[18F]cGBP4F	low yield	low	slower	moderate

Overall, these tracers provide novel useful neuroimaging tools for studying/detecting neuropathic pain in the laboratory by autoradiography and biodistribution. Even though [³H]gabapentin is commercially available and can potentially be used for autoradiography and biodistribution studies, it is often not practical to conduct these studies with tritiated drugs as in vivo studies require very large amounts of tritiated drug, autoradiography may take weeks and biodistribution using liquid scintillation counting requires homogenization, bleaching of the tissues and tissue-specific calibration curves. Future efforts to develop tracers for α2δ-1 receptor should focus on tracers with higher nerve uptake and low background signal. In addition, although the focus of this work was not to develop a brain imaging agent, future directions should include evaluating these agents for brain imaging and developing gabapentin tracers with high brain uptake as those may have application for imaging CNS diseases such as epilepsy and psychiatric conditions. Finally, although these tracers did not allow detecting injured nerves in rats due to the small size of the nerves and the signal in surrounding muscle tissue, these molecules might be useful in larger animals or humans. In many chronic pain patients, it is not possible to identify if their pain is neuropathic in origin and these tools may assist in identifying such a component, whether it changes over time and may contribute to selection of the most appropriate therapeutic intervention.

Data availability

Data will be made available on request.