Synthesis and biological evaluation of pyrrolidine-based T-type calcium channel inhibitors for the treatment of neuropathic pain

Abstract The treatment of neuropathic pain is one of the urgent unmet medical needs and T-type calcium channels are promising therapeutic targets for neuropathic pain. Several potent T-type channel inhibitors showed promising in vivo efficacy in neuropathic pain animal models and are being investigated in clinical trials. Herein we report development of novel pyrrolidine-based T-type calcium channel inhibitors by pharmacophore mapping and structural hybridisation followed by evaluation of their Cav3.1 and Cav3.2 channel inhibitory activities. Among potent inhibitors against both Cav3.1 and Cav3.2 channels, a promising compound 20n based on in vitro ADME properties displayed satisfactory plasma and brain exposure in rats according to in vivo pharmacokinetic studies. We further demonstrated that 20n effectively improved the symptoms of neuropathic pain in both SNL and STZ neuropathic pain animal models, suggesting modulation of T-type calcium channels can be a promising therapeutic strategy for the treatment of neuropathic pain.


Introduction
Neuropathic pain is chronic pain originated from lesions or diseases of the peripheral or central somatosensory nervous system 1 . It is usually associated with unpleasant feeling such as a burning, abnormal crawling, or electric shock-like sensations and is further characterised by evoked pain after stimuli such as hypersensitivity and summation. Current therapeutic options for neuropathic pain include anti-convulsants such as sodium and calcium channel blockers, anti-depressants, opioids, NSAIDs, cannabinoids, topical agents, and combinations of these drugs 2,3 . However, administration of these drugs only reduce 30-50% of pain in approximately 50% of neuropathic pain patients 4,5 . This low efficacy issue can be attributed to the substantial heterogeneity of pain mechanisms, and requests precise elucidation of underlying mechanisms for different types of neuropathic pain in order to improve both diagnosis and prescription. Thus, identification of potential biomarkers for each type of neuropathic pain and subsequent development of chemical probes that can modulate the biomarkers will provide exciting opportunities to better understand the heterogeneity of neuropathic pain and produce efficacious pain drugs 6,7 .
One of the encouraging strategies to treat neuropathic pain is modulation of intracellular calcium levels 8,9 either directly by controlling voltage-gated calcium channels or indirectly by regulating receptors such as nicotinic acetylcholine receptors 10,11 . Among several voltage-gated calcium channels, T-type calcium channels have received great attention as one of the promising molecular targets for neuropathic pain 12,13 . These are low voltage-activated (LVA) channels composed of a1 subunit single pore, and three isoforms of T-type calcium channels have been identified 14 : Ca v 3.1 (a 1G ) 15 , Ca v 3.2 (a 1H ) 16 , Ca v 3.3 (a 1I ) 17 . In contrast to high voltageactivated (HVA) channels, T-type calcium channels are activated even after a slight depolarisation of the cell membrane, functioning at near-resting membrane potentials 18 . Due to this unique voltage sensitivity, T-type calcium channels can regulate neuronal excitability and oscillatory behaviour 19 .
Among T-type calcium channels, Ca v 3.2 is the major isoform expressed in primary sensory afferent neurons and spinal cord 20 , and it has been strongly implicated that Ca v 3.2 channel contributes to the development of neuropathic pain 12 . For example, in a neuropathic pain rat model derived by a chronic constriction injury (CCI), the expression of Ca v 3.2 channel is highly up-regulated, and delivery of Ca v 3.2-specific antisense oligonucleotide generated robust pain-relieving effects 20 . Furthermore, in a STZ-induced neuropathic pain model, the level of Ca v 3.2 expression is also increased and specific in vivo knock-down of Ca v 3.2 effectively alleviated mechanical and thermal hypersensitivity 21 . In addition, Ca v 3.1 channels are mainly expressed in dorsal horn neurons 22 , and its significance in neuropathic pain is also well reported. It was demonstrated that after spinal nerve ligation (SNL) Ca V 3.1 knock-out mice showed a higher threshold to mechanical stimuli than their wild-type litter mates, and thermal hypersensitivity was also substantially decreased 23 . Furthermore, the mechanical hypersensitivity was markedly attenuated after induction of trigeminal neuropathy in the Ca V 3.1 knock-out mice as compared to wild type mice 24 .
Based on these observations, it was believed that T-type calcium channel inhibitors would provide effective treatment options for neuropathic pain 13 , leading to development of a variety of T-type calcium channel blockers, such as Mibefradil 25 , ethosuximide 26 , and (3b,5a,17b)-17-hydroxyestrane-3-carbonitrile (ECN) 27 . In fact, treatment of Mibefradil or ethosuximide in the rat CCI model relieved behavioural symptoms of neuropathic pain 28 . It was shown that ECN also alleviated mechanical and thermal hypersensitivity in rats with neuropathic pain 29 and in diabetic ob/ob mice 30 . Although these early T-type channel inhibitors showed great promise for the treatment of neuropathic pain, however, they suffered relatively low selectivity against other ion channels, especially voltage-gated sodium channels in neurons raising adverse side effect issues 25,31,32 . Moreover, Mibefradil was withdrawn just 1 year after the FDA approval due to drug-drug interactions 33 .
Thus, development of selective T-type channel inhibitors is highly desired to treat neuropathic pain with minimal side effects. Recently, selective T-type channel inhibitors such as R-(-)-efonidipine 34 , TTA-P2 35 , ML-218 36 , ABT-639 37 , Z-944 38 , and benzodiazepine-and dihydropyrazole-based Actelion inhibitors 39,40 have been developed, and some of them presented robust analgesic effects in several neuropathic pain models 41,42 , illustrating selective T-type calcium channel inhibitors hold strong potential to be effective treatment options for neuropathic pain.
Herein we describe our ongoing efforts to develop selective T-type channel inhibitors, identifying pyrrolidine derivatives as potent T-type channel inhibitors after evaluation of Ca v 3.1 (a 1G ) and Ca v 3.2 (a 1H ) T-type calcium channel blocking activities with fluorescence-based FDSS6000 assay 43 and whole-cell patch clamp recordings 44 . Pharmacokinetic parameters of a promising compound 20n and its in vivo analgesic efficacy in two different neuropathic pain models are investigated.

Whole-cell patch-clamp method
Ca v 3.1: Ca v 3.1 channel was stably expressed in HEK293 cells. The cells used for T-type calcium channel activity assay were cultured on a coverslip coated with poly-L-lysine (0.5 mg/mL) whenever sub-cultured, and their calcium channel activity was recorded 2-7 days after the cultivation. Current of the T-type calcium channel at a single cell level was measured according to an electrophysiological whole-cell patch-clamp method using EPC-9 amplifier (HEKA, Germany). At this time, a cell exterior solution composed of 140 mM NaCl, 2 mM CaCl 2 , and 10 mM HEPES (pH 7.4) and a cell interior solution composed of KCl 130 mM, HEPES 10 mM, EGTA 11 mM, and MgATP 5 mM (pH 7.4) were employed. Inward current caused by the T-type calcium channel activation which occurred when the cells were converted into a whole-cell recording mode by stabbing a microglass electrode having 3-4 MX resistance which was filled with the cell interior solution into a single cell and depolarised at -30 mV (50 ms duration period) every 10 s with fixing membrane potential to -100 mV was measured according to a T-type calcium channel protocol activated at low current. Ca v 3.2: current of the calcium channel expressed from the Xenopus unfertilised oocytes was measured according to a two-electrode voltage clamp method. Current was measured in 10 mM Ba 2þ solution composed of 10 mM Ba(OH) 2 , 90 mM NaOH, 1 mM KCl, 0.1 mM EDTA, and 5 mM HEPES, and pH was adjusted to 7.4 with methanesulfonic acid. At this time, voltage clamp and current measurements were regulated with an amplifier for unfertilised oocytes (Model OC-725C, Warner Instrument Corp., LLC., Hamden, CT, USA), analog signals were converted into digital signals using Digidata 2000 A (Analog-Digital converter, Axon Instrument), and acquisition, storage, and analysis of all data were recorded in Pentium IV computer via pCLAMP8. The data were mainly collected at 5 KHz and filtered at l KHz (Model 902 filter; Frequency devices located at Haverhill, MA, USA). The generation of T-type current was occurred by imposing test electric potential of -20 mV every 15 s on the unfertilised oocytes whose potential was fixed at -90 mV, and a blocking percentage was calculated by comparing the potentials before and after the drug treatment.
In vivo assay Behavioural tests were conducted for rats at 1 day prior to surgery and 14 days after surgery. After the postoperative behavioural test, the animals were treated orally with 100 mg/kg compound 20n or gabapentin. The tests were reevaluated at 1 h, 3 h, and 5 h after administration. Mechanical allodynia: Testing was performed according to methods described in the literature 45 . Rats were acclimated in a transparent plastic boxes permitting freedom of movement with a wire mesh floor to allow access to the planter surface of the hind paws. A von Frey filament (Stoelting, Wood Dale, IL) was applied 5 times (once every 3-4 s) to hind paw. Von Frey filaments were used to assess the 50% mechanical threshold for paw withdrawal. The 50% withdrawal threshold was determined by using the up-down method and formula given by Dixon 46 : 50% threshold ¼10 (X þ kd) /10 4 , where X is the value of the final von Frey hair used (in log units), k is the tabular value for the pattern of positive/negative responses modified from the same literature, and d is the mean difference between stimuli in log units (0.17). Cold allodynia: To quantify cold sensitivity of the paw, brisk paw withdrawal in response to acetone application was measured as reported previously 47 . The rat was placed under a transparent plastic box on a metal mesh floor and acetone was applied to the plantar surface of the hind paw. To do this, an acetone bubble was formed at the end of a small piece of polyethylene tubing which was connected to a syringe. The bubble was then gently touched to the heel. The acetone quickly spread over the proximal half of the plantar surface of the hind paw. The acetone was applied 5 times to the hind paw at 2 min interval. The frequency of paw withdrawal was expressed as a percentage [(no. of trials accompanied by brisk foot withdrawal/total no. of trials) Â l00]. The results of behavioural tests are expressed as a %MPE. For example, paw withdrawal thresholds were converted to %MPE by the following formula, using a cutoff of 15 g (the threshold for normal rats) to define maximum possible effect: (post drug threshold-baseline threshold)/(cutoff-baseline threshold) Â 100. %MPE values near 100 indicate normal mechanical thresholds (i.e. at or near 15 g), whereas values near 0 indicate allodynia. The result of cold allodynia was also expressed as %MPE.

Inhibitor design
Many promising T-type channel inhibitors have been developed based on a piperidine moiety such as Penfluridol (1) 48 , TTA-P2 (2) and Z-944 (3) (Figure 1). We also synthesised T-type channel inhibitors based on a piperidine moiety such as compound 4 which showed approximately 60% inhibition of Ca v 3.1 at 10 lM 49 . However, pyrrolidine, a close homolog of piperidine, has not been frequently introduced so far. Therefore, in an effort to develop novel potent T-type channel inhibitors, we decided to synthesise pyrrolidine-based inhibitors. Our initial molecular structure was constructed via a structure-hybridisation between the template of TTA-P2 or Z-944 and 5-isobutyl-1-phenyl-1H-pyrazole moiety that consistently showed robust activity in our previous T-type inhibitors ( Figure 2) 50,51 . Furthermore, this compound series accommodating hydrophobic R group are well mapped with 3D ligand-based pharmacophore that was produced by common feature hypothesis generation approach (HipHop) implemented in CATALYST program for the rational design of T-type calcium channel inhibitors (Figure 3) 52 .

Biological activity
In vitro T-type calcium channel inhibitory activity The in vitro T-type calcium channel inhibitory activity of synthesised pyrrolidine compounds were evaluated against Ca V 3.1 and  Ca V 3.2 at two different concentrations, 10 lM and 1 lM, respectively (Table 1). We stably expressed Ca V 3.1 and Ca V 3.2 in HEK293 cells and measured the concentration of intracellular Ca 2þ with fluo3/AM after the activation of Ca V 3.1 and Ca V 3.2 channels by 70 mM of KCl in 10 mM of CaCl 2 . Then, we presume that the degree of the fluorescence decrease read by FDSS6000 after coincubation with synthesised compounds can reflect the inhibitory activity of each pyrrolidine compound 43 . First, the free pyrrolidine 15 was subjected to this FDSS6000-based fluorescent assay, but the compound was not able to inhibit both T-type channels more than 20% at 10 lM (18.60% inhibition against Ca V 3.1 and 10.72% inhibition against Ca V 3.2). Guided by the pharmacophore mapping results we decided to put a hydrophobic group to the free pyrrolidine to increase the potency. N-Alkylation of pyrrolidine dramatically increased inhibitory activities as exemplified with 16 and 20a which inhibited the activation of Ca V 3.1 and Ca V 3.2 channels more than 47% at 10 lM. Since 20a was more potent than 16, we focused on the synthesis of N-phenethylpyrrolidine compounds. A variety of substituents such as methyl-, fluoro-, chloro-, trifluoromethyl-and trifluoromethoxy groups were introduced to the phenyl ring to investigate the substituent effects on the activity. Most compounds showed great inhibitory activities more than 65% against both channels at 10 lM, and 10 compounds (20b-h, 20 m, 20n, 20p) inhibited greater than 80% of activation of at least one  of either Ca V 3.1 or Ca V 3.2 channels. Furthermore, IC 50 values of selected compounds were precisely measured by the whole-cell patch-clamp assay 44 , and compounds displaying high inhibition percentage at 10 lM also showed great IC 50 values against both T-type calcium channels (IC 50 (Ca V 3.1) 2.14-6.11 lM, IC 50 (Ca V 3.2) 2.88-8.14 lM) strongly implicating that these pyrrolidine compounds possess great therapeutic potential for neuropathic pain.
In vitro ADME properties In order to further evaluate the potential of these pyrrolidine compounds as promising therapeutic agents, their respective physicochemical properties were evaluated by assessing CYP450 inhibitory activities at 10 lM and metabolic stability in human liver microsome ( Table 2). Most compounds barely inhibited CYP1A2 enzyme, but several compounds such as 20k showed moderate to strong inhibition of other tested CYP enzymes. Also, we determined remaining compound percentage at 30 min after incubation of compounds in human liver microsome, but unfortunately the metabolic stability of synthesised pyrrolidines appeared not quite high. Nevertheless, among N-phenethylpyrrolidine compounds, 20n seemed most stable in human liver microsome, approximately 15% of 20n left intact after 30 min. As a next step, IC 50 values of selected compounds against hERG channels were evaluated employing the same whole-cell patch-clamp assay ( Table 2). Blockage of hERG channels induces long QT syndrome leading to potentially fatal cardiac arrhythmia 54 . Often, t-type calcium channel inhibitors suffer from low selectivity against hERG channels which we should investigate in developing t-type channel blockers to avoid potential cardiac side effects. The IC 50 value of Mibefradil against hERG channels was reported as approximately 1.34 lM 49 . It was observed that most of our compounds showed lower IC 50 values than Mibefradil requiring further improvement of selectivity against hERG channels. Fortunately, better selectivity against hERG channels than Mibefradil was obtained with compounds 16, 20k and 20n, among which 20n exhibited IC 50 value of 8.13 lM representing safest compounds in this series of compounds. Comprehensive evaluation of current pyrrolidine compound series based on T-type channel inhibitory activity and in vitro ADME properties led us to determine 20n as a promising candidate for the next in vivo evaluation.
In vivo pharmacokinetic study First, in vivo pharmacokinetic studies were carried out with Sprague-Dawley (SD) male rats (Table 3). Total eight rats were equally divided into two groups and 10 mg/kg of 20n was administered to each group but in two different routes, orally and intravenously. Then, pharmacokinetic parameters of each administration route were calculated from compound concentrations in the blood. It turned out moderate C max value (0.20 lg/mL) was observed after the oral administration. Nevertheless, because the clearance rate (29.7 ml/min/kg) was relatively low, total AUC 0-1 values after both oral and intravenous dosing were 339.17 and 61.89 lg min/mL sufficient enough to exert therapeutic effects in vivo. Of particular importance is the observation that brain-to-plasma ratio at 2 h was 8.78 after intravenous administration, strongly suggesting that once exposed to the plasma, 20n is highly likely to cross the tight blood-brain barrier efficiently and well penetrated into the brain. Despite low oral bioavailability presumably caused by poor metabolic stability, 20n can be employed to develop promising candidates for the treatment of neuropathic pain.
In vivo efficacy studies with two neuropathic pain animal models In order to further demonstrate the potential of 20n as a therapeutic agent to treat neuropathic pain, we tested in vivo efficacy of 20n with two different animal models of neuropathic pain: the SNL model 55 and the streptozotocin (STZ)-induced diabetic neuropathy model 56 . The SNL model is generated by ligating one (L5) or two (L5 and L6) segmental spinal nerves of the rat. In addition, streptozotocin is known to destroy pancreatic beta cells via DNA alkylation followed by activation of apoptosis resulting in diabetes. Both models result in behavioural symptoms such as increased sensitivity to non-noxious mechanical or cold stimuli and heat hypersensitivity, and have been widely used as animal model systems to investigate neuropathic pain. In the SNL model, we first observed that the paw withdrawal threshold to mechanical stimuli was significantly decreased and also the frequency of the paw Table 2. Results of in vitro ADME assay (CYP inhibition, microsomal stability, hERG inhibition) of pyrrolidine-based inhibitors 16 and 20a-p. withdrawal to cold stimuli was substantially increased confirming that neuropathic pain was induced 2 weeks after the surgery ( Figure 4). Then, 20n was orally administered to rats (n ¼ 4) with 100 mg/kg. The current drug for neuropathic pain, gabapentin (100 mg/kg) was also dosed orally to another rat group (n ¼ 4) side by side to directly compare pain-relieving effects with 20n. The mechanical threshold was evaluated by von Frey filaments. It was found that 20n was able to restore mechanical hypersensitivity close to the normal level and showed superior analgesic effects as compared to gabapentin at 3 h after the treatment. Furthermore, 20n significantly lowered the paw withdrawal response rate to cold stimuli as well, and the analgesic effects of 20n were again similar to that of gabapentin.
After observing treatment of 20n relieved neuropathic pain response in the SNL model, we decided to further prove that the ability of 20n to reduce neuropathy can be general and are not confined to one specific neuropathic model. The streptozotocin (STZ)-induced diabetic neuropathy model for diabetic peripheral neuropathy (DPN) was selected. We confirmed that intraperitoneal injection of 65 mg/kg of streptozotocin to rats reduced both the paw withdrawal threshold to non-noxious mechanical stimuli and the paw withdrawal latency to heat stimuli at 2 weeks after the surgery ( Figure 5). Subsequently, 100 mg/kg of 20n and gabapentin was orally dosed to seven and six rats, respectively. Once again, 20n increased the paw withdrawal threshold similarly to gabapentin, and delayed the paw withdrawal latency comparable Table 3. Pharmacokinetic parameters of 20n after intravenous and oral administration (10 mg/kg) to SD male rats. Values are presented as mean ± SD. AUC: area under the plasma concentration versus time curve); C max : peak plasma concentration; T max : time to reach C max ; CL: time-averaged body clearance; MRT: mean residence time; V ss : apparent volume of distribution at steady state; F: oral bioavailability. a Median (range) for T max . to gabapentin suggesting 20n is also effective in alleviating heat hypersensitivity.

Conclusion
In conclusion, we have developed pyrrolidine-based T-type calcium channel inhibitors by pharmacophore mapping and structural hybridisation followed by evaluation of their Ca v 3.1 and Ca v 3.2 channel inhibitory activities with FDSS and whole-cell patch-clamp assay. Most pyrrolidine compounds potently inhibited activation of both Ca v 3.1 and Ca v 3.2 T-type channels showing IC 50 values between 2.14 and 8.14 lM. Profiling of in vitro ADME properties revealed that among this series 20n has great selectivity against hERG channel and high plasma exposure in the subsequent in vivo pharmacokinetic study. Furthermore, it was notable that 20n can be potentially well penetrated into the brain according to the excellent brain to plasma compound concentration ratio at 2 h after intravenous administration. Finally, 20n exhibited comparable analgesic effects to gabapentin in the both SNL and STZ neuropathic pain animal models, implicating potent T-type calcium channel inhibitor 20n can be developed as promising candidates to treat neuropathic pain. Further optimisation of 20n in regard to physicochemical properties like metabolic stability is currently in progress.