Structure−Activity Relationship Studies of Tetrahydroquinolone Free Fatty Acid Receptor 3 Modulators

Free fatty acid receptor 3 (FFA3, previously GPR41) is activated by short-chain fatty acids, mediates health effects of the gut microbiota, and is a therapeutic target for metabolic and inflammatory diseases. The shortage of wellcharacterized tool compounds has however impeded progress. Herein, we report structure−activity relationship of an allosteric modulator series and characterization of physicochemical and pharmacokinetic properties of selected compounds, including previous and new tools. Two representatives, 57 (TUG-1907) and 63 (TUG-2015), showed improved solubility and preserved potency. Of these, 57, with EC50 = 145 nM and a solubility of 33 μM, showed high clearance in vivo but is a preferred tool in vitro. In contrast, 63, with EC50 = 162 nM and a solubility of 9 μM, showed lower clearance and seems better suited for in vivo studies. Using 57, we demonstrate for the first time that FFA3 activation leads to calcium mobilization in murine dorsal root ganglia.


■ INTRODUCTION
Short-chain fatty acids (SCFAs) are produced in large amounts by the lower gut microbiota and are known to affect human health in various and often beneficial ways. 1 Free fatty acid receptors 2 and 3 (FFA2 and FFA3) are G protein-coupled receptors activated by SCFAs and mediate many of the physiological effects of SCFAs. 2,3 The receptors were discovered and reported simultaneously in 2003 and are co-expressed in enteroendocrine cells, pancreatic β-cells, some immune cells, and certain cancers. 4−7 Both FFA2 and FFA3 have been reported to be expressed in the adipose tissue, although most studies now suggest that only FFA2 is present. 6−10 Of the two receptors, FFA2 has received more attention, showing promise as a target for the treatment of various metabolic and inflammatory conditions, 6 with one compound reaching clinical trials for ulcerative colitis before being discontinued due to limited efficacy, despite the fact that the compound did inhibit neutrophil infiltration. 11 Several studies have also suggested interesting therapeutic potential for FFA3. For example, Offermanns and co-workers demonstrated that deletion of FFA2 and FFA3 in combination, but not individually, improved insulin secretion and glucose tolerance in mice, indicating that dual antagonism of the receptors may counteract type 2 diabetes. 12 Activation of both receptors is also reported to counteract cancer development. 13,14 FFA3 is linked to hypoxia-induced apoptosis and may have potential as a target for ischemia/reperfusion-related injury. 15 Marsland and co-workers found that FFA3 but not FFA2 mediates the protective effect of circulating SCFAs against allergic lung inflammation and is therefore of interest for treatment of allergic asthma. 16 FFA3 has also been reported in both autonomic and somatic sensory ganglia. 17 In sympathetic ganglia, propionate is found to promote sympathetic nervous system activation and to be involved in regulation of the body energy balance. 18 Most of the studies involving FFA3 have relied on SCFAs as tools and/or knockout mice. These studies should be interpreted with caution because SCFAs are generally able to activate both FFA2 and FFA3 but with different profiles on human and rodent orthologues 19 and because it has been found that knockout of one receptor may affect the expression of the other. 10 More studies are therefore required to elucidate the therapeutic potential of FFA3 alone and in combination with FFA2. An important reason for the paucity in studies on FFA3 is the lack of well-characterized high-quality tool compounds for this receptor. 20 Studies that use SCFAs often employ propionate (C3) as a dual agonist of FFA2 and FFA3, acetate as a FFA2selective agonist, and butyrate as a FFA3-selective agonist; however, the selectivity for these compounds are modest at best. 19 Recently, the FFA3-selective SCFA-analogue 1-methylcyclopropylcarboxylate (1-MCPC) has also been employed, but its potency remains very low. 21 Currently, the only tools for FFA3 with potency in the singledigit micromolar range are from a series of tetrahydroquinolones originally disclosed by Arena Pharmaceuticals (represented by 1, Table 1) and that were subsequently shown to be allosteric modulators of the receptor. 22,23 Although they have occasionally been used as tools, 24−28 moderate potency and the lack of proper characterization have limited the use of these compounds. The compounds act as allosteric modulators and are not affected by mutation of arginine residues in the orthosteric site that are indispensable to the activity of propionate. 22 Small structural changes of the tetrahydroquinolones have shown to affect the mode of action, ranging from pure allosteric agonists to modulators that either enhance the potency of propionate (positive allosteric modulators, PAMs), reduce the efficacy of propionate (negative allosteric modulators, NAMs), or acting as both agonists on their own but also enhance the potency of propionate (PAM agonists).
Herein, we present the results from a thorough examination of the structure−activity relationships (SAR) within this compound series and further characterize bioavailability and pharmacokinetic properties of the most promising analogues. Moreover, we use a key compound to demonstrate the functional activity of FFA3 in cells of murine dorsal root ganglia.

■ SYNTHESIS
The tetrahydroquinolone target compounds were typically synthesized from the appropriate 3-ketoamide, aldehyde, and 3-aminoenone using the Hantzsch dihydropyridine synthesis (Scheme 1). Heating at 80°C in i-PrOH for up to 5 days generally gave the best outcome, with longer reaction times for more hindered substrates. 29 Microwave heating or synthesis from 3-ketoamide, aldehyde, 1,3-dione, and ammonia provided the product in shorter time but at the expense of lower yield and purity. The 3-ketoamide substrates were most conveniently accessed by heating of the appropriate aniline in neat methyl acetoacetate. These intermediates were also synthesized by heating of the aniline with 2,2,6-trimethyl-4H-1,3-dioxin-4-one or in the presence of Lewis acid catalysts but generally with an inferior outcome. The preferred route to the typical tetrahydroquinolone is shown in Scheme 1.

■ RESULTS AND DISCUSSION
The ligands were initially screened in a human FFA3-dependent [ 35 S]GTPγS binding assay as this assay reflects receptormediated activation of G i/o proteins and is known to generally correlate well with ligand affinity. 30 Analogues of particular interest were tested further in a cAMP inhibition assay as this is an important downstream effect of G i -activation. The latter was applied as a standard assay because of higher reproducibility. It is furthermore performed in whole cells, is more downstream, includes G protein signal amplification, and therefore better reflects a more natural ligand−receptor response. Because the series binds to an allosteric site on FFA3, 22 selected compounds were also tested together with a fixed submaximal concentration of the SCFA propionate to investigate potential allosteric effects on orthosteric agonist function.
Tetrahydroquinolones 1 and 2 were disclosed by Arena Pharmaceuticals in 2006. 23 We resynthesized these compounds, confirming FFA3 agonist activity in the low micromolar range. 22 2-Furyl derivative 1 showed agonist activity with approximately 2-fold higher potency than 3-furyl derivative 2 in the GTPγS assay (Table 1). Replacing the 2-furyl with phenyl (3) led to significant deterioration of potency, whereas 2-thienyl (4), a group with a size and polarity that more closely resembles phenyl than furyl, largely preserved potency in the GTPγS assay.
The analogue with 2-bromophenyl (5) was previously characterized and found to be inactive alone but to act as a PAM of propionate, implying that 5 binds to FFA3 but is unable to activate the receptor directly. 22 Introducing bromosubstituents in the 3-(6) or 4-position (7) of phenyl derivative 3 regained most of the activity relative to 2-furyl derivative 1 in the GTPγS assay, whereas the potency was essentially constant for 3−4 and 6−7 in the cAMP assay. Analogues with 3-(8) and 4trifluoromethyl (9) and 4-methyl (10) substituents were full agonists with potencies similar to or lower than that of the unsubstituted 3, whilst extension to 4-ethyl (11) eroded potency.
Replacement of o-tolyl by phenyl (12) gave a >3-fold drop in potency, whereas m-tolyl (13) or p-tolyl (14) resulted in a more moderate drop, indicating the ortho-position as the most interesting. Introduction of methoxy (15) further eroded potency in the GTPγS binding assay but retained potency in the cAMP inhibition assay. The 2,5-dichloro-substituted AR420626 (16) also originates from Arena Pharmaceuticals and has been described as a tool compound in the literature and characterized by us. 22,24,25 Like 1, this compound exhibited only moderate potency in the GTPγS assay 22 but is an order of magnitude more potent in the cAMP assay. It thus represents one of the most potent compounds but is also known to have poor solubility.
We also wished to explore aliphatic R 1 groups, and we were pleased to find that isobutyl (17) behaved as an FFA3 agonist with a potency similar to that of 1 in the [ 35 S]GTPγS binding assay. Introducing 2,3-dimethyl (18), 2-iodo (19), 2-chloro (20), or 2,6-difluorophenyl (21) on the R 2 phenyl while keeping R 1 as isobutyl produced active compounds with potencies comparable to that of 17, with 20 representing an improvement in the cAMP but not in the GTPγS assay. In general, results from the cAMP assay corresponded satisfactorily with the GTPγS data, although some of the compounds deviated considerably. Notably, increased potency of 10-, 14-, and 50-fold was observed for 2, 3, and 15, respectively. In contrast, only 17 and 18 exhibited a lower potency of 5-and 2-fold, respectively. On the other hand, together with 1 μM propionate, 17 exhibited a potency that was equal with 1 (with or without propionate). As the cAMP assay is more downstream and showed a reproducibility that was at least as good as the GTPγS assay, this was chosen as the primary assay for the remaining compounds.
We next turned our attention to the other parts of the structure (Table 2). An extension of the ortho-methyl at the dihydropyridine to ethyl (22) led to an order of magnitude decrease of potency while a phenyl (23) produced a completely inactive compound. Oxidation of the dihydropyridine to pyridine (24) also produced an inactive compound, perhaps unsurprisingly because this compound represents substantial structural changes.
Using 17 as a starting point, opening of the cyclohexanone and formation of a phenone (25) or methyl ester (26) both produced compounds that were inactive (pEC 50 < 4) alone but Scheme 1. General Synthetic Route for Tetrahydroquinolone Ligands Table 2. Scaffold Exploration of Furyl and Isobutyl Analogues a 26 acted as PAM with propionate, indicating that some larger changes in the structure also produced compounds with affinity for FFA3. Introduction of a nitrile (27) produced a compound that was active alone, albeit with low potency, but that was substantially potentiated by the presence of propionate. We also wished to explore the significance of the ketone; however, all attempts to reduce the cyclohexanone carbonyl or derivatize, for example, to form oximes, were unsuccessful, partly due to low electrophilicity due to stabilization by conjugation to the enamine of the dihydropyridine system and partly due to instability of products.
Replacement of the anilide part by a methyl ester (28) or a phenone (29) produced compounds that also were inactive alone but exhibited PAM properties. Introduction of a carboxylic acid in this position produced a compound that could not be characterized or tested due to insufficient solubility.
We next reverted to the R 1 group to further explore aliphatic substituents ( Table 3). Introduction of sec-butyl (30) resulted in a compound with very similar properties to isobutyl 17. The sterically more well-defined cyclohexyl (31) gave marginally increased potency which could be a result of a hydrophobic effect. Surprisingly, the slightly smaller cyclopentyl (32) increased potency 20-fold to 288 nM. Cyclopropyl (33) maintained good potency at 600 nM with an order of magnitude reduced lipophilicity. A series of n-alkyls from ethyl to pentyl (34−37) resulted in agonists with micromolar potency, with propyl (35) and butyl (36) being the most potent and the shorter ethyl (34) the least potent. Further extension with phenethyl (38), styryl (39), and pyrazolylethyl (40) continued this trend with pEC 50 values <5, interestingly, with the less lipophilic 40 exhibiting the highest potency of the three. All compound were full agonists.
Identification of the cyclopentyl as the most potent aliphatic substituent motivated another venture into similarly sized aromatic substituents. N-Methyl-2-pyrazole (41) was a lowpotency compound, in line with ortho-substituted phenyls such as 5. Furyls and thienyls with small substituents resulted in better potency, with 5-bromo-2-furyl (44) and 2-benzofuryl (45) as the most potent. The lower potency of 3-benzothienyl 46 is likely due to the positioning of the benzene ring.
In an attempt to increase aqueous solubility by decreasing lipophilicity, the 2-furyl of 16 was replaced by 2-thiazolyl (47), 5-thiazolyl (48), and 4-thiazolyl (49) ( Table 3). This strategy failed, as the solubility dropped from 5 to below 2 μM for the thiazolyl analogues in a kinetic solubility assay, although the potency was only slightly decreased for 47 (Table 4). Also, even though clogP of 47−49 indicated an order of magnitude improvement of lipophilicity relative to 16 (log D 7.4 = 3.19), measured log D 7.4 revealed similar or increased lipophilicity. Replacement by 4-oxazolyl (50) improved solubility >10-fold but had a detrimental effect on potency. On the other hand, 2oxazolyl (51) improved potency but reduced solubility to 1 μM. With a focus on solving the solubility problem, we returned to the aliphatic substituents aiming to incorporate a positive charge. Unfortunately, both small polar and larger, lipophilic amine substituents (52−54) produced inactive compounds. Analogues with negatively charged carboxylate groups in the same position (not shown) were also explored but were completely inactive. Boc-protected intermediates (55−56) were tested and found to be inactive alone. 52−55 were also inactive in the presence of 3 μM propionate and were therefore also not PAMs. However, 56, representing the compound with the largest R 1 substituent explored in this study, turned out to be a NAM, effectively inhibiting receptor signaling at 30 μM concentration ( Figure 1). This is in line with previous compounds characterized as FFA3 NAMs, where R 1 was 3-or 4-phenoxyphenyl and substantially larger than other R 1 -groups explored in the series. Thus, a larger R 1 increases the chance of finding a NAM or antagonist.
We reasoned that the generally poor physicochemical properties of the 2,5-dichlorophenyl compounds are worsened with aromatic or larger, lipophilic R 1 -groups. With the failure of positively charged R 1 groups to produce FFA3 agonists, we therefore proceeded with exploration of smaller neutral groups. The combination of n-propyl with 2,5-dichlorophenyl (57) indeed produced a compound with similar potency to 16 but 6fold increased solubility (Table 5). Hoping to further increase solubility, we next investigated the requirements of chlorinated R 2 -groups in relation to potency. The 2-chloro (58) or 5-chloro   (59) substituents alone were accompanied by 4−5-fold increased solubility compared to 57 but reduced potency down to a level similar to o-tolyl (35, Table 3). The solubility of 35 was 199 μM and its log D 7.4 was 2.64 and thus favorable compared to all chlorinated compounds and comparable with 1 (190 μM solubility, log D 7.4 2.05). A tendency toward higher potency for 58 indicated that the 2-chloro is more important than the 5-chloro, in line with observations for methyl substituents (cf. 1, 13, 14, Table 1). In an attempt to regain potency but keep the solubility properties, the 5-chloro of 57 was replaced by methyl (60) or methoxy (61). This indeed resulted in compounds with solubilities similar to the other monochlorinated compounds, but the potency was also similar or only marginally improved. Chemical stability was also tested, and all compounds (58−61) were completely recovered after 1 week at 37°C in PBS. Finally, revisiting small aliphatic cycles, cyclopropyl (62) partly regained and cyclopentyl (63) fully regained the potency of the 2-furyl analogue 16, but with similar or only moderately increased solubility. Compounds 62 and 63 were also tested in the cAMP assay together with 1 μM propionate but did not reveal significant PAM effects. Overall, of the compounds tested as PAMs, none with pEC 50 > 6 showed significantly enhanced potency in the presence of propionate.
To identify the most active enantiomer, racemic 1 was resolved by chiral HPLC and crystallized. The single crystal X-ray structures for one enantiomer ( Figure 2) along with that of the racemate ( Figure S1) were determined. The absolute configuration of the most active enantiomer (R)-1 was determined by anomalous dispersion effects with a Flack parameter of 0.04 (12). However, the assignment must be viewed with some caution because of the high standard deviation in the Flack parameter. Crystal packing is influenced by the presence of a single or both enantiomers in the lattices of (R)-1 and (R,S)-1, respectively. Both structures show that the strongest intermolecular interaction between molecules is the H-bonding between the cyclic carbonyl (O1) and the amine NH group (N1) of the adjacent molecule ( Figure S2). These Hbonds link the molecules in ribbons approximately parallel to the a-and c-axis in the structures of (R)-1 and (R,S)-1, respectively ( Figures S5 and S2). The ribbons stack with the 2-furyl rings located on the same side of every molecule in (R)-1 ( Figure S6). In (R,S)-1, they are alternating in the up and down positions in accordance with the H-bonded ribbons comprising alternating R and S enantiomers ( Figure S3).
The previously published compound 1 was not affected by mutation of either R185A or R258A in the orthosteric binding site, and the compounds were therefore believed to be allosteric modulators. In the search for the potential allosteric binding site, we mutated a third arginine residue in a neighboring site (R71A); however, this did not affect the potency of 1. After the  Journal of Medicinal Chemistry pubs.acs.org/jmc Article publication of the FFA1 crystal structure in complex with an allosteric agonist located in the TM region close to the intracellular site, 32 a homology model of hFFA3 was constructed using Modeller. The model revealed a similar potential binding site on FFA3 and initial docking studies of (R)-1 indicated three amino acids that might be involved in binding ( Figure 3a). Mutation of E112A and R126A did not affect the potency of (R)-1 or propionate, whereas Q131A significantly reduced the potency of (R)-1 (>3-fold, p < 0.001) but did not affect the potency of propionate. Thus, additional docking of (R)-1 in this site were performed to find poses where Q131, but not R126, show a central role in binding. Seven out of 10 poses showed a hydrogen bond interaction between Q131 and either the amide carbonyl or the ketone without any constraints used and only one of the poses showed an additional hydrogen bonding to R126 ( Figure 3b). The narrow binding cavity around the core scaffold is in agreement with the restrained SAR observed for this series and the furan moiety pointing out of the binding cavity might explain the flexibility in substituents at this site ( Figure 3c,d). Although Q131A impacts the potency of (R)-1 with high significance, the magnitude of the effect is lower than what would be generally expected for the removal of a hydrogen bond interaction; thus, additional studies are required to confirm the proposed binding site.
With several compounds with improved properties in hand, we wished to investigate the suitability of the compounds as tools for in vivo studies in rodents. Because of the moderate potency and low solubility associated with this series, it has been presumed that they are poorly suited as in vivo tools. A basic requirement is preserved activity on the relevant species orthologues, which generally has been observed to be low within the free fatty acid receptor group. 19 We confirmed that propionate maintained potency between human and rodent species. 1 and 16 both lost 3−10-fold potency from human to rodent species (1 pEC 50 = 5.87 ± 0.07 on rFFA3, 5.42 ± 0.04 on mFFA3; 16 pEC 50 = 6.34 ± 0.04 on rFFA3, 5.88 ± 0.05 on mFFA3), whereas 63 now tended toward the highest potency on the rodent orthologues (pEC 50 = 6.39 ± 0.03 on rFFA3; pEC 50 = 5.96 ± 0.04 on mFFA3).
We next investigated the chemical stability of the compounds and stability toward liver microsomes. Representative compounds were shaken in PBS at pH 7.4. Apart from some that precipitated, all compounds were quantitatively recovered after 1 week. On the other hand, the tested compounds exhibited varying stability towards mouse liver microsomes (MLM, Table  6). In the one end, 1 and 63 showed good stability, comparable with the propranolol reference compound (61%). In the other end, only 1−2% was recovered of 16, 47, 57, and 62. The remaining compounds were in the intermediate range. Apart from a generally higher stability of the o-tolyl compounds 1 and 35 than the remaining 2,5-dichlorophenyl compounds, it was difficult to see a clear relationship between structural or physicochemical properties and microsomal stability.
Based on the results so far, 63 appeared as a good compromise between the properties of 1 and 16, and 57 represented a good compromise between potency and solubility. We therefore selected these four compounds for pharmacokinetic studies in mice. Although the microsomal data for at least 16 and 57 indicated that these compounds would have a fast clearing, we decided to run the pharmacokinetic study using the same time frame for all compounds. Overall, the four compounds had surprisingly favorable PK properties and especially 1, 16, and 63 showed high bioavailability. Compound 1 exhibited the longest half-life, approximately an hour after iv dosing, and the lowest clearance, whereas 16 showed a half-life of 20 min. 63 was somewhat in between the two, and 57 had the shortest iv halflife. Half-life after oral dosing was satisfactory for all compounds. 1, 16, 57, and 63 were counterscreened on the related FFA receptors FFA1 and FFA2, the more distant FFA receptors FFA4 and GPR84, and the L-type calcium ion channels, a target for related dihydropyridine ligands. No significant activity was detected at any of the receptors at 10 μM concentration (see Supporting Information).
Because of its good compromise between solubility and potency (EC 50 = 2 μM at mFFA3 vs 30 μM for propionate in the GTPγS assay), 57 was considered for further in vitro studies. To further ensure the suitability of 57 as a tool compound for in vitro studies, the chemical stability was also evaluated in dimethyl sulfoxide (DMSO). An NMR sample was stored at 4°C , and recordings after 1 and 30 days showed complete stability. A sample stored at rt (not protected from light) was fully stable for 7 days, whereas minor decomposition was detected after 30 days (see Supporting Information).
Thus, 57 was selected for studies in cells from dorsal root ganglia (DRGs) isolated from wild-type and FFA3 KO mice. Addition of 57 (5 μM) gave a large elevation in intracellular calcium levels in wild type but not FFA3 KO cells, indicating a FFA3-specific effect of 57 on DRG cells (Figure 4). These results suggest, in line with previous studies, 17,18 a key role for FFA3 in mediating effects for SCFAs from the gut microbiome within the peripheral nervous system. They furthermore for the first time demonstrate a pharmacological intervention at ganglionic FFA3, that 57 is able to evoke a response with high selectivity and that ganglionic FFA3 has potential as a therapeutic target for pain and metabolic disorders. Further

■ CONCLUSIONS
We have performed an extensive SAR investigation around a series of 1,4,7,8-tetrahydroquinol-5-one FFA3 allosteric modulators that has led to the identification of key structural parts required for inducing effects on FFA3 and parts where modifications are permitted. The studies revealed that the aromatic substituents at the 4-position (referred to as R 1 ) can be replaced by small aliphatic substituents with fully preserved potency. Extension of the R 1 -group leads to loss of agonist activity but may produce antagonists or NAMs. The 2,5dichlorophenyl at the amide (R 2 ) has a beneficial effect on potency but a detrimental effect on solubility. The combination of N-2,5-dichlorophenyl with a smaller aliphatic substituent in the 4-position on the tetrahydroquinol-5-one resulted in compounds with preserved potency and improved solubility. The most potent enantiomer was found to have the (R)configuration. We have for the first time evaluated the in vitro solubility, lipophilicity, and metabolic stability and in vivo pharmacokinetic properties of members of this compound series. Despite largely suboptimal result from in vitro studies, the compounds overall demonstrated surprisingly favorable in vivo pharmacokinetic properties. Compound 63 was found to have well preserved potency on rodent orthologues, good stability toward MLMs, and favorable PK properties for use as an in vivo tool compound. The previously characterized agonists 1 and 16 were also found to have surprisingly favorable PK profiles, although the former suffers from moderate potency and the latter from low solubility. Compound 57 was found to have sufficiently good aqueous solubility and good potency but showed the least favorable PK properties of the studied compounds; thus this is the preferred compound for in vitro but not in vivo studies.  33 the structures were solved with the ShelXT 34 structure solution program using Intrinsic Phasing and refined with the ShelXL 34 refinement package using Least Squares minimization. All nonhydrogen atoms were refined using anisotropic atomic displacement parameters, and hydrogen atoms were inserted at calculated positions using a riding model, except those belonging to the secondary amine and the amide groups. These hydrogen atoms were located in different electron density maps, and their positions were refined. The methyl group C22 in (R,S)-1 is disordered over both ortho positions on the aromatic ring, and the occupancy was refined to 76%:24% occupancy. Crystal data for (R,S)-1: C 22  General Tetrahydroquinoline (THQ) Procedure. A vial was charged with 3-aminocyclohex-2-en-1-one (1 equiv.), aldehyde (1−1.2 equiv.), 3-oxo-N-(o-tolyl)butanamide (1 equiv.), and isopropanol (IPA) (5 mL/mmol). The vial was capped and heated to 80°C for 1−5 days. Afterward, the reaction mixture was cooled to room temperature, diluted with EtOAc, and concentrated on Celite before purification by flash chromatography.
Dorsal Root Ganglion Assay. Colonic innervating DRGs were isolated from the T9-L2 region of the spinal cord of C57/BL6 mice and from FFA3 knock-out animals and immediately placed in cold Hanks' balanced salt solution (HBSS; Sigma-Aldrich). Isolated DRGs were initially digested with HBSS containing L-cysteine (0.3 mg/mL) and papain (2.0 mg/mL) for 20 min at 37°C. The solution was removed and replaced with HBSS contain collagenase (4.0 mg/mL) and dispase (4.0 mg/mL) (20 min at 37°C) for further digestion. The collagenase solution was then replaced with DMEM to stop the reaction. The DRGs were finally dissociated by mechanical trituration using a pipette. Dissociated cells were plated on matrigel-coated coverslips and placed in an incubator (37°C and 5% CO 2 ). Following a 2 h incubation, cells were flooded with 90% DMEM (Sigma) supplemented with 10% fetal calf serum and 1% PenStrep and further incubated overnight at 37°C and 5% CO 2 .
To measure intracellular calcium and its potential regulation, dissociated cells on the coverslips were loaded with Fura 8-AM (3 μM) (Stratech Scientific Limited) for 20 min at 37°C in the dark. Coverslips were then placed in a recording chamber and mounted onto an inverted fluorescent microscope (Nikon TE2000-E; Nikon Instruments, Melville, NY) equipped with a (NA = 1.3) oil-immersion Super Fluor objective lens (×40), an Optoscan monochromator (Cairn Research, Faversham, Kent, UK), and a digital Cool Snap-HQ CCD camera (Roper Scientific/Photometrics, Tucson, AZ). Illumination of the preparation was achieved by a Meta Fluor imaging software (Molecular Devices, San Jose CA, version 7.8.8).
Cluster of cells were randomly selected for real time imaging and continuously perfused with HEPES buffer (composition: HEPES 10 mM, NaCl 135 mM, glucose 10 mM, KCl 5 mM, CaCl 2 2 mM and MgCl 2 1 mM, pH 7.4) for 20 min at room temperature. All test ligands were diluted in HEPES buffer and perfused through the chamber for 3 min, followed by a final application of the Ca 2+ ionophore ionomycin (5 μM), as a positive control.