Functional Selectivity of 6′-Guanidinonaltrindole (6′-GNTI) at κ-Opioid Receptors in Striatal Neurons*

Background: 6′-Guanidinonaltrindole (6′-GNTI) activates G protein coupling to κ-opioid receptors (KOR) without β-arrestin2 recruitment in transfected cells. Results: In striatal neurons, 6′-GNTI activates Akt but not ERK1/2; U69,593 activates both kinases. Conclusion: In neurons, U69,593-induced activation of ERK1/2 is β-arrestin2-dependent, whereas activation of Akt is G protein-mediated. Significance: Identification of KOR signaling pathways in endogenous systems will inform the development of KOR-directed medications. There is considerable evidence to suggest that drug actions at the κ-opioid receptor (KOR) may represent a means to control pain perception and modulate reward thresholds. As a G protein-coupled receptor (GPCR), the activation of KOR promotes Gαi/o protein coupling and the recruitment of β-arrestins. It has become increasingly evident that GPCRs can transduce signals that originate independently via G protein pathways and β-arrestin pathways; the ligand-dependent bifurcation of such signaling is referred to as “functional selectivity” or “signaling bias.” Recently, a KOR agonist, 6′-guanidinonaltrindole (6′-GNTI), was shown to display bias toward the activation of G protein-mediated signaling over β-arrestin2 recruitment. Therefore, we investigated whether such ligand bias was preserved in striatal neurons. Although the reference KOR agonist U69,593 induces the phosphorylation of ERK1/2 and Akt, 6′-GNTI only activates the Akt pathway in striatal neurons. Using pharmacological tools and β-arrestin2 knock-out mice, we show that KOR-mediated ERK1/2 phosphorylation in striatal neurons requires β-arrestin2, whereas Akt activation depends upon G protein signaling. These findings reveal a point of KOR signal bifurcation that can be observed in an endogenous neuronal setting and may prove to be an important indicator when developing biased agonists at the KOR.

physiological functions. In particular, the dynorphin/KOR system has been shown to modulate dopaminergic and serotonergic tone in the CNS and therefore has become a favorable target for drug discovery efforts aimed at treating addiction and mood disorders. Current KOR ligands are limited by unwanted side effects and unusual pharmacodynamics. KOR agonists, for example, have been shown to decrease the reinforcing and rewarding effects of abused drugs (1)(2)(3)(4)(5) and reduce the symptoms of mania (6,7). However, they also induce sedation, dissociation, and dysphoria (8). KOR antagonists have antidepressant and anxiolytic effects in rodent models (9 -11) and may provide a therapeutic means of treating relapse to addictive drugs (12,13); yet the currently available, selective KOR antagonists are extremely long acting, which may limit their clinical utility (14,15). Therefore, the question remains as to how this receptor should be acted upon to mediate the desired effects while limiting undesired consequences, whether that be through the use of partial agonists or functionally selective agonists or antagonists.
Functional selectivity, biased agonism, and ligand-directed signaling conceptually describe the condition wherein ligand binding to G protein-coupled receptors (GPCR) promotes receptor coupling to certain signaling cascades preferentially over others (16 -20), and evidence of ligand-directed signaling at the KOR has been reported. For example, U69,593 and MOM-salvinorin B differ in their ability to induce astrocyte proliferation and in the temporal activation of ERK1/2 observed in astrocytes following drug treatment. In these cells, U69,593 activates ERK1/2 via both early phase G protein-dependent and later ␤-arrestin2-dependent mechanisms, the latter of which was correlated with astrocyte proliferation. In contrast, MOM-salvinorin B only activates the early G proteindependent ERK1/2 pathway and does not induce astrocyte proliferation (21). In another demonstration of functional selectivity at KOR, the classic KOR antagonists norbinaltorphimine (nor-BNI) and JDTic were shown to antagonize G protein coupling but activate the MAP kinase JNK in a KOR-mediated manner (22). It is currently hypothesized that functional selec-tivity results from ligands directly influencing the conformation of the receptor and thereby directing selective interactions with particular second messenger systems. When coupled with molecular modeling studies and the introduction of the recent crystal structure of KOR, the characterization of KOR ligands that promote distinct pharmacological profiles may lead to structure-function-inspired drug design (23)(24)(25).
6Ј-Guanidinonaltrindole (6Ј-GNTI) was developed by Portoghese and colleagues (26) as a guanidino derivative of the selective ␦-opioid receptor (DOR) antagonist naltrindole. Interestingly, the addition of a guanidino group to naltrindole converts it to a KOR-selective compound, and moving the guanidino group from the 5Ј to 6Ј position changes the ligand from a KOR antagonist (5Ј-GNTI) to a KOR agonist (6Ј-GNTI). Initially, 6Ј-GNTI agonist activity at KOR was demonstrated by showing that 6Ј-GNTI-induced inhibition of smooth muscle contraction in the guinea pig ileum could be selectively blocked by nor-BNI (26). Recently, 6Ј-GNTI was reported to partially activate G protein-mediated signaling pathways without inducing ␤-arrestin2 recruitment in HEK-293 cells (27). These findings make 6Ј-GNTI of great interest for dissecting the roles of different KOR signaling cascades as well as KOR drug development for the treatment of pain and addiction. In the present study, we sought to further characterize the pharmacological properties of 6Ј-GNTI in multiple signaling pathways in both KOR-expressing Chinese hamster ovary (CHO) cells and primary neuronal cultures where the KOR is expressed endogenously. We confirm that in CHO cells, 6Ј-GNTI is a potent partial agonist in G protein signaling and that it antagonizes ␤-arrestin2 recruitment to the KOR and receptor internalization (27). Although 6Ј-GNTI does not recruit ␤-arrestins, it does induce the phosphorylation of ERK1/2 and Akt, suggesting that these kinases are activated independent of KOR/␤-arrestin2 interactions. However, when we assessed kinase activation profiles in primary striatal neurons, we found that only U69,593 is able to induce ERK1/2 activation and it does so in a ␤-arrestin2-dependent and pertussis toxin-insensitive manner. Conversely, both U69,593 and 6Ј-GNTI induce Akt phosphorylation in striatal neurons, activation that is blocked by pretreatment with pertussis toxin. These results indicate that 6Ј-GNTI is a functionally selective ligand at the endogenously expressed KOR.

EXPERIMENTAL PROCEDURES
Cell Culture and Cell Line Construction-CHO-K1 cells were transfected with an N-terminal hemagglutinin (HA)-tagged human KOR in pcDNA3.1(ϩ) (Gene ID 4986, Missouri S&T cDNA Resource Center, Rolla, MO) by electroporation (GenePulser XCell, Bio-Rad). Transfected cells were selected via HA-antibody cell sorting by flow cytometry (FACSAria and LSRII, BD Biosciences), and the top 3% expression level cell population was used for all experiments. CHO-KOR cells were characterized for KOR membrane expression via confocal microscopy and total receptor number determined by [ 3 H]U69,593 saturation binding (K D ϳ 0.6 nM, Bmax ϳ 1800 fmol/mg protein) (28). For all experiments with the CHO-KOR line, untransfected CHO cells served as negative controls (data not shown).
Drugs-U69,593 was obtained from Sigma, and nor-BNI dihydrochloride, 5Ј-GNTI ditrifluoroacetate, and 6Ј-GNTI ditrifluoroacetate were all obtained from Tocris Bioscience (Ellisville, MO). U69,593 was prepared in ethanol, 6Ј-GNTI and 5Ј-GNTI were prepared in DMSO, and nor-BNI was prepared in dH 2 O. In all cases, vehicle contained 1% DMSO and 1% ethanol in PBS (unless stated otherwise in the figure legends). Pertussis toxin from Bordetella pertussis was obtained from Sigma.
Mice-␤-Arrestin2 knock-out (␤arr2-KO) and wild-type (WT) littermates were used to assess [ 35 S]GTP␥S coupling in mouse brain and for the culturing of primary neurons (28). All experiments were performed with the approval of the Institutional Animal Care and Use Committee of The Scripps Research Institute.
[ 35 S]GTP␥S Coupling-CHO-KOR cells were collected, and membrane pellets were prepared by homogenization with a polytronic Tissue-Tearor in buffer (10 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM EDTA) followed by 20,000 ϫ g centrifugation at 4°C for 30 min. The resultant membrane pellet was resuspended in assay buffer (50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 5 mM MgCl 2 , 1 mM EDTA, 40 M GDP) via Teflon-onglass Dounce homogenizer. Fifteen g of membrane protein was incubated with increasing concentrations of drug and 0.1 nM [ 35 S]GTP␥S (PerkinElmer Life Sciences) for 1 h at 30°C in a total reaction volume of 200 l. For antagonist experiments, protein was preincubated with test compounds for 15 min prior to the addition of 100 nM U69,593 and [ 35 S]GTP␥S. Reactions were terminated by rapid filtration using a 96-well plate Brandel cell harvester (Brandel, Gaithersburg, MD) followed by washes with ice cold water. Microscint-20 (PerkinElmer Life Sciences) was added to the plates after drying, and radioactivity was read with a TopCount NXT HTS microplate scintillation and luminescence counter (PerkinElmer Life Sciences). All compounds were run in parallel assays in duplicate for comparison.
For coupling in mouse brain, striata were isolated from 4 -5month-old male WT mice. Tissue was homogenized, and membranes were prepared as described above. Coupling reactions, containing 2.5 g of protein, 20 M GDP, and 0.1 nM [ 35 S]GTP␥S, were incubated at room temperature for 2 h prior to harvesting, as described above.
␤-Arrestin2 Translocation-To visually assess ␤-arrestin2 translocation to the KOR, CHO-KOR cells were transiently transfected with ␤arr2-GFP (5 g) by electroporation and plated on collagen-coated glass-bottom confocal dishes (MatTek, Ashland, MA). After incubation at 37°C for 24 h, the cells were serum-starved for 1 h in serum-free MEM without phenol red (Invitrogen). Drugs were added, and live cell images were obtained at the time points indicated using an Olympus FluoView IX81 confocal microscope (Olympus, Center Valley, PA) as described previously (29).
␤-Arrestin2 translocation was also studied using the PathHunter ␤-arrestin assay in CHO-K1 cells expressing the KOR (DiscoveRx, Fremont, CA) according to the manufacturer's protocol. Cells were treated with agonist for 90 min prior to assessment of enzyme complementation. For antagonist experiments, the cells were incubated with the antagonist for 60 min prior to agonist addition. Luminescence values were obtained using a Synergy HT luminometer (BioTek, Winooski, VT). All compounds were run in parallel experiments in duplicate.
KOR Internalization-To qualitatively determine agonist-induced receptor internalization, CHO-KOR cells were plated on collagen-coated glass-bottom confocal dishes and serumstarved for 30 min prior to the experiment. The KOR was labeled in live cells by incubation with an anti-HA-AlexaFluor 488 antibody (1:100, Invitrogen) for 15 min at 37°C. Following washes, drug (10 M) was added, and the cells were monitored by confocal imaging for HA-KOR internalization over the times indicated (29,30).
Agonist-induced KOR internalization was also assessed in CHO-KOR cells using a cell surface biotinylation assay. This assay was performed as described previously (30). The cells were serum-starved for 30 min, and surface receptors were biotinylated prior to drug treatment. Drug was added at 10 M, and cells were treated for 30 min or 1 h (30 min for the antagonist and then 30 min for the agonist, where applicable). The remaining surface biotinylation was removed by glutathionestripping buffer, and the protected biotinylation was immunoprecipitated by streptavidin-coated beads. Western blots were performed as described previously with 10% bis-Tris NuPage gels (Invitrogen) and transferred to nitrocellulose membranes (18,30). The receptor was detected using a KOR antibody (Neuromics, Edina, MN). Blots were imaged using the Odyssey infrared imager, and densitometry of bands was measured using Odyssey 2.1 software (Li-Cor, Lincoln, NE). The -fold stimulation over vehicle was calculated by dividing the densitometry values of the treated groups by the average value of the vehicle-treated group.
In-cell Western-To assess agonist-induced ERK1/2 phosphorylation via In-cell Western assay (Li-Cor), CHO-KOR cells were plated into 384-well clear-bottom, black-walled plates (BD Biosciences) at a density of 15,000 cells/well. The cells were serum-starved for 1 h prior to treatment with increasing drug concentrations for 10 min. The cells were fixed (4% paraformaldehyde in PBS), washed in 0.1 M glycine, permeabilized (0.2% Triton X-100), and blocked with Li-Cor blocking buffer in PBS (Li-Cor). Cells were then incubated with both phosphorylated and total ERK1/2 primary antibodies (p-ERK1/2, catalog No. 4370, at 1:300, and t-ERK1/2, catalog No. 4696, at 1:200; Cell Signaling Technology, Beverly, MA) overnight in Li-Cor blocking buffer. Secondary antibodies (anti-rabbit IRDye800 and anti-mouse IRDye680, both at 1:500; Li-Cor) were added in Li-Cor blocking buffer ϩ 0.025% Tween 20. Fluorescence was determined using the Odyssey infrared imager. The -fold ERK1/2 phosphorylation over vehicle levels was determined by first normalizing phosphorylated ERK1/2 levels to total ERK1/2 levels within each well and then normalizing to the average of the vehicle treated cells within each experiment. All compounds were run in parallel with 6 replicate wells per point.
Signaling by Western Blot-CHO-KOR cells were plated in 6-well dishes and serum-starved for 1 h in serum-free medium prior to treatment. Concentrations and durations of treatment are indicated in the legends for Figs. 4 and 5. For antagonist studies, nor-BNI was added during the last 15 min of the serum starvation. For pertussis toxin pretreatment studies, 100 ng/ml pertussis toxin was added to the complete medium at ϳ24 h prior to the assay. Following treatment, cells were washed once in PBS, and lysates were prepared in lysis buffer (20 mM Tris, 150 mM NaCl, 2 mM EDTA, 0.1% SDS, 1% Nonidet P-40, 0.25% deoxycholate, 1 mM sodium orthovanadate, 1 mM PMSF, 1 mM NaF, and protease inhibitors). Western blots were performed to assess phospho-ERK1/2 (1:1000, Cell Signaling Technology, catalog No. 4370), total ERK1/2 (1:1000, Cell Signaling Technology, catalog No. 4696), phospho-Akt (1:1000, Cell Signaling Technology, catalog No. 2965), and total Akt (1:1000, Cell Signaling Technology, catalog No. 2920) levels as described above.
Striatal Neurons-Primary striatal neuronal cultures were generated from postnatal day 1 mouse neonates obtained from homozygous breeding of ␤arr2-KO and WT mice. Striata were isolated and digested in 0.0375% papain (Sigma-Aldrich) followed by trituration to generate a single cell suspension as described previously (18,31). Neurons were plated in Neurobasal-A medium (Invitrogen) supplemented with 2% B-27 serum supplement, 0.5 mM L-glutamine, 1.675 g/liter sodium selenite, 2.5 mg/liter insulin, 1.375 mg/liter transferrin, and 50 g/ml gentamycin onto 12-well tissue culture dishes coated with poly-L-lysine (Sigma-Aldrich). Cytosine ␤-D-arabinofuranoside (Sigma-Aldrich) was added to the culture medium 4 days after plating, and the medium was replaced with fresh culture medium at 5 days post-plating.
On the tenth day after plating, kinase activation was determined. Neurons were incubated in serum-free MEM for 1 h prior to drug treatment. For agonist studies, neurons were treated with 10 M drug for 10 min. For nor-BNI pretreatment studies, nor-BNI was added to the medium to a final concentration of 1 M during the last 15 min of the serum starvation period. For pertussis toxin pretreatment studies, pertussis toxin was added to the medium 24 h prior to drug treatment at a final concentration of 100 ng/ml. Following drug treatment, neurons were washed once in PBS; lysates were prepared, and Western blots were performed as described above.
Data Analysis and Statistics-All statistical comparisons were made using GraphPad Prism 6.01 software (GraphPad, La Jolla, CA). To generate concentration response curves, vehicle values were subtracted from the raw data, and three-parameter nonlinear regression curve-fit analysis was performed. For determination of agonist stimulation, the percentage of maximal U69,593 was calculated by normalizing all values to the top of the U69,593 curve. For antagonist studies, values were normalized to the average of 100 nM or 1 M U69,593 alone, and the percentage is provided (Figs. 1B and 2C). The potency and efficacy values, obtained from the averages of the nonlinear regression analysis performed on each individual curve, are reported as the mean Ϯ S.E. in Tables 1 and 2. To determine ligand differences within and between assays, curves were fit to the operational model (32), and the ratio of /K A was calculated (GraphPad Prism). The within assay differences were determined by subtracting the log(/K A ) for U69,593 (the reference ligand) from the values obtained for 6Ј-GNTI, which generated the "transduction efficiency" for 6Ј-GNTI, or the ⌬log(/K A ). To determine bias between assays, relative bias factors were calculated by subtracting the ⌬log(/K A ) for each assay, presented as ⌬⌬(/K A ) path1-path2 (Table 1) (33-35). Student's t test indicates an unpaired, two-tailed analysis. In all cases, signifi-cance was considered to be p Յ 0.05. All studies were performed n Ն 3 independent experiments performed in multiple replicates, with the specific number of individual experiments for each assay noted in the figure legends.

RESULTS
G protein coupling assays were used to compare 6Ј-GNTI and the reference agonist, U69,593. 6Ј-GNTI proved to be more potent that U69,593, although it behaved as a partial agonist ( Fig. 1A and Table 1). As expected, the antagonists had no effect. The classification of 6Ј-GNTI as a partial agonist was further confirmed by its partial inhibition of U69,593-stimulated G protein coupling, as compared with 5Ј-GNTI and nor-BNI ( Fig. 1B and Table 2).

TABLE 1 Agonist activity at KOR in the CHO-KOR cell line
Emax is expressed relative to maximal U69,593-induced activation. The EC 50 and Emax values (mean Ϯ S.E., n Ͼ 3) were calculated by nonlinear regression analysis compared with the reference ligand, U69,593. The change in transduction efficiency (⌬TE) within each assay was calculated based on the equation Individual n values are presented in the figure legends. Student's t test analysis, U69,593 vs. 6Ј-GNTI: *, p Ͻ 0.05; ***, p Ͻ 0.001. 5Ј-GNTI and Nor-BNI failed to converge to stimulatory nonlinear regression curve-fitting analysis. The transduction ratio between assays ⌬⌬log(/K A ) G protein-␤arr2 was calculated; the divergence from 1 indicates the degree of bias for G protein coupling over ␤-arrestin2 recruitment for 6Ј-GNTI relative to the performance of the reference ligand, U69,593, in each assay (33,34).  vented and reversed U69,593-induced ␤arr2-GFP translocation (Fig. 2, D and E).

Ligand
6Ј-GNTI was also tested for its ability to induce KOR internalization, which has been shown to be a ␤-arrestin-dependent process (36). Confocal images reveal that the N-terminal HAtagged KOR translocates from the surface of CHO cells into intracellular vesicles following U69,593, but not 6Ј-GNTI, treatment (Fig. 3A). U69,593-induced internalization of the KOR can also be blocked by pretreating the cells with either 6Ј-GNTI or 5Ј-GNTI (Fig. 3B). Results from a cell surface biotinylation assay support the findings of the confocal studies, wherein treatment with U69,593, but not 6Ј-GNTI or 5Ј-GNTI, induces KOR internalization, and both 6Ј-GNTI and 5Ј-GNTI fully block U69,593-induced internalization (Fig. 3C). These findings demonstrate, in both a different cellular background and in different signaling assays than those previously reported (27), that 6Ј-GNTI maintains its bias toward G protein-mediated signaling over ␤-arrestin2 recruitment.
To assess the impact of the 6Ј-GNTI bias toward the activation of G protein coupling over ␤-arrestin2 recruitment on the activation of downstream signaling, ERK1/2 and Akt phosphorylation profiles were assessed in the CHO-KOR cell line. Both 6Ј-GNTI and U69,593 robustly induce ERK1/2 phosphorylation, whereas 5Ј-GNTI has no effect when assessed in an immunocytochemistry-based assay. In this assay, 6Ј-GNTI was again more potent than U69,593 and yet only slightly less efficacious ( Fig. 4A and Table 1). ERK1/2 activation was confirmed by Western blot analysis, and the effects of both drugs can be antagonized by nor-BNI (Fig. 4B). As KOR-induced ERK1/2 phosphorylation can be biphasic, with an early and late phase of activation (21), we compared the time courses of kinase activation for the 10 M concentration of U69,593 and 6Ј-GNTI by Western blot analysis. U69,593 and 6Ј-GNTI induced similar temporal profiles of ERK1/2 activation, with phosphorylation levels peaking by 10 min and persisting over basal levels for up to 2 h post-drug treatment (Fig. 4C). Pertussis toxin pretreatment completely abolished both U69,593-and 6Ј-GNTI-mediated ERK1/2 activation, indicating a G␣ i/o -dependent signaling mechanism (Fig. 4D). 6Ј-GNTI and U69,593 stimulated Akt phosphorylation in a very similar manner to that observed for ERK1/2 (Fig. 5, A-C). Again, Akt phosphorylation in CHO-KOR cells can be blocked by pertussis toxin pretreatment (Fig. 5D).
Both the ligand and the complement of intracellular proteins available to interact with a receptor can dictate the pathway by which a receptor signals, and therefore ligand signaling profiles can vary in different cell types (19,37). To assess whether the ligand bias we observed between U69,593 and 6Ј-GNTI is maintained in a more physiologically relevant system, we assessed the ability of 6Ј-GNTI to activate KOR-mediated signaling pathways in striatum and in primary striatal neurons from mice. In striatal membranes prepared from adult mice, U69,593 induced stimulation of G protein coupling, whereas 6Ј-GNTI had no effect (Fig. 6). It is common to encounter small assay windows when performing G protein coupling in brain tissue, as evidenced by the 40% stimulation achieved by the full agonist, U69,593. Therefore, it is not surprising that 6Ј-GNTI, which could only stimulate 35% of the maximal coupling achieved with U69,593 in the overexpression system, did not produce detectable stimulation of G protein coupling in this system. To overcome this limitation of the system, we tested whether 6Ј-GNTI would maintain its role as a partial agonist by determining whether it could displace U69,593-stimulated coupling. In accordance with its role as a partial agonist at KOR, 6Ј-GNTI, when included at a constant concentration, suppressed the efficacy of U69,593, demonstrating that 6Ј-GNTI has a direct impact on KOR-mediated G protein coupling in striatum.
In striatal neurons cultured from WT neonates, U69,593 induced the phosphorylation of ERK1/2, which was blocked by pretreatment with nor-BNI (Fig. 7A). In contrast to U69,593, exposure to 6Ј-GNTI caused a decrease in ERK1/2 phosphorylation to a level below that observed in vehicle treated neurons. The 6Ј-GNTI-induced repression of ERK1/2 was not sensitive to nor-BNI pretreatment; however, nor-BNI alone suppressed ERK1/2 phosphorylation below vehicle levels, suggesting that 6Ј-GNTI may act through a mechanism similar to nor-BNI to dampen the basal activation of this pathway. The time course of 6Ј-GNTI-induced ERK1/2 phosphorylation was not shifted, as treatment with 6Ј-GNTI for 2 min also revealed no ERK1/2 phosphorylation, demonstrating that the kinase activation window was not simply overshot by the 10-min incubation period.
Interestingly, when we assessed Akt phosphorylation in WT striatal neurons following treatment with U69,593 and 6Ј-GNTI, we found that both agonists induced Akt phosphor-

DISCUSSION
In comparison with U69,593, 6Ј-GNTI is a highly potent, partial agonist in G protein coupling (Fig. 1) as well as in ERK1/2 (Fig. 4) and Akt (Fig. 5) phosphorylation assays when assessed in KOR-expressing CHO cell lines. We confirmed here that 6Ј-GNTI promotes only marginal ␤-arrestin2 recruitment in the commercial DiscoveRx PathHunter TM cell line and no detectable recruitment of a GFP-tagged ␤-arrestin2 to KOR, as assessed by confocal microscopy (Fig. 2). Consistent with its weak ability to recruit ␤-arrestin2, 6Ј-GNTI did not promote subsequent receptor internalization (Fig. 3). Overall, 6Ј-GNTI displayed signaling bias toward G protein coupling over ␤-ar-restin2 recruitment in the cell lines (Table 1).
When the two agonists are compared for induction of KOR signaling in neuronal cultures, it was found that U69,593 activates ERK1/2, whereas 6Ј-GNTI does not. Conversely, 6Ј-GNTI decreases the overall level of ERK1/2 activation, an effect that is also observed with the KOR antagonist nor-BNI (Fig. 7). It is attractive to hypothesize that the lack of ERK1/2 activation is due to the inability of 6Ј-GNTI to recruit ␤-arres-tin2 and that ␤-arrestin2 may mediate KOR signaling to ERK1/2. This idea is supported by the following observations: 1) U69,593-induced ERK1/2 activation is not sensitive to pertussis toxin, suggesting that KOR to ERK1/2 signaling is G protein-independent; and 2) U69,593-induced ERK1/2 signaling does not ensue in the absence of ␤-arrestin2, as demonstrated in striatal neurons isolated from ␤arr2-KO mice.
membranes supports this idea. However, treatment of striatal neurons with 6Ј-GNTI stimulated Akt phosphorylation to the same extent as treatment with U69,593, suggesting that the agonist is capable of fully activating KOR in this setting (Fig. 8). Moreover, the activation of Akt is pertussis toxin-sensitive, suggesting that 6Ј-GNTI, as well as U69,593, utilizes a G␣ i/o protein signaling pathway to induce this response in striatal neurons. It is not clear, however, why 6Ј-GNTI behaves as a full agonist for activating AKT in the neurons when it only partially activates G␣ i/o coupling in brain. This could be due to the differences in brain tissue (whole striatum) versus neuronal culture preparations. Alternatively, it could reflect differences in assessing the proximal G protein coupling in the isolated membranes versus measuring the accumulation of the downstream amplified AKT phosphorylation that is measured after treating live, intact neurons. Another potential explanation is that the G protein-mediated signal to Akt is less hampered when the ␤-ar-restin2 is antagonized by 6Ј-GNTI, as suggested in the cell culture assays.
Whereas KOR-mediated activation of ERK1/2 requires ␤-ar-restin2 in striatal neurons, ERK1/2 phosphorylation in CHO cells appears to be completely dependent upon pertussis toxinsensitive G protein signaling. It is not clear from this study why the neurons and the CHO cells differ regarding their utilization of ␤-arrestin2 in this signaling cascade; however, such observations have precedence. For example, serotonin, N-methyltryptamine, and 2,5-dimethoxy-4-iodoamphetamine induce Akt phosphorylation downstream of the serotonin 2A receptor in C6 glioma and PC12 cells (38), 6 but only serotonin induces Akt phosphorylation in mouse cortical neurons, a process that has been shown to require ␤-arrestin2 to mediate serotonininduced activation of Akt in mouse cortical neurons (18). It is attractive to speculate that there are components of the immediate environment of KOR endogenously expressed in neurons, that are not present in the transfected Chinese hamster ovary cells and that provide an assembly of a functional "receptorsome" unique to the striatal neurons. These observations suggest that KOR expressed in striatum require ␤-arrestin2 to activate ERK1/2. The nature of these neuronal culture-specific complements remains to be determined, as well as the physiological and behavioral consequences of such signaling. The development and evaluation of agonists that selectively activate one pathway over another, much like 6Ј-GNTI, will allow for such elucidations. Overall, the findings described here underscore the importance of evaluating compound functional selectivity in an endogenous setting.
In the current study, 6Ј-GNTI was shown to be very potent at activating KOR-mediated G protein coupling, which is in agreement with a study showing that 6Ј-GNTI potently inhibits cAMP accumulation (27). However, in HEK-293 cells expressing a modified G␣ protein (⌬6-Gqi4-myr) coupled to a calcium channel reporter assay, the potency of 6Ј-GNTI was substantially lower (ϳ224 nM) than the potency observed for U69,593 (ϳ0.8 nM) (39). The co-expression of the DOR along with the KOR improved the potency of 6Ј-GNTI (ϳ40 nM) but did not affect the potency for U69,593, which led us to conclude that 6Ј-GNTI is a weak agonist at KOR but a more potent agonist at the KOR/DOR heteromer. This observation was not recapitulated in studies assessing cAMP accumulation, where DOR coexpression had no effect on the potency of 6Ј-GNTI (27). In the 6 C. L. Schmid and L. M. Bohn, unpublished observations. FIGURE 8. U69,593-and 6-GNTI-mediated Akt phosphorylation in mouse primary striatal neurons is pertussis toxin-sensitive and ␤-arrestin2-independent. Primary striatal neurons were serum-starved for 1 h prior to the determination of Akt phosphorylation (10 M for 10 min). Representative images of Western blots and densitometric analyses are provided. For each sample, phosphorylated Akt (p-Akt) was first normalized to total Akt (t-Akt), and the -fold stimulation over vehicle is provided (mean Ϯ S.E.). A, WT neurons were pretreated with either vehicle (Veh, 0.9% saline) or nor-BNI (Nor, 1 M) during the last 15 min of the serum starvation. Akt is activated by both U69,593 (U69) and 6Ј-GNTI (6Ј) in a nor-BNI-sensitive manner (one-way ANOVA: for U69,593, F (2,13) ϭ 8.247, p ϭ 0.0049; for 6Ј-GNTI, F (2,14) ϭ 5.662, p ϭ 0.0158; Bonferroni's post hoc test: vehicle versus drug treatment groups, *, p Ͻ 0.05, **, p Ͻ 0.01; drug versus nor-BNI ϩ drug, , p Ͻ 0.05; n ϭ 5-6 independent neuronal preparations). Nor-BNI treatment alone does not alter Akt phosphorylation from vehicle levels (Student's t test: p Ͼ 0.05; n ϭ 4 independent neuronal preparations). B, U69,593-and 6Ј-GNTI-induced Akt activation in WT striatal neurons is sensitive to overnight pretreatment with pertussis toxin (PTX; 100 ng/ml) (Student's t test: vehicle (V) versus drug treatment within each pretreatment group, **, p Ͻ 0.01; n ϭ 3 independent neuronal preparations). C, U69,593 and 6Ј-GNTI induce an increase in Akt phosphorylation levels in striatal neurons cultured from ␤arr2-KO mice (Student's t test: ***, p Ͻ 0.001; n ϭ 4 independent neuronal preparations). present study, we also found that 6Ј-GNTI is a potent, partial agonist at the KOR using a standard [ 35 S]GTP␥S binding assay in CHO-KOR cellular membranes, and as only KOR is overexpressed in the cell line, presumably the actions due to activation of KOR. One potential explanation for the discrepancies is the use of different cell lines. Another possible explanation may be found in the nature of the assays themselves. In contrast to the [ 35 S]GTP␥S binding assay, which assesses interactions between the receptor and its endogenous G proteins, the ⌬6-Gqi4-myrexpressing cell line measures changes in calcium fluorophore release due to receptor activation of a chimeric G␣ q/i protein (40). Therefore, these differences may not be reflective of the low capacity of 6Ј-GNTI to activate the KOR monomer/homomer but rather its relative low potency in promoting KOR coupling to the modified, chimeric G proteins. Collectively, these findings emphasize the importance of the cellular context in receptor and drug function, as the complement of intracellular proteins, which differ between cell types (i.e. the presence of homomers or heteromers, G protein-coupled receptor kinases, and other scaffolding or regulatory proteins), will determine the signaling and regulatory events that occur downstream of receptor activation.
Discrepancies between pharmacological characterizations made across cell-based assays and between those made in endogenous cell types underscore the need to be cautious in assigning classifications of ligands (such as agonist, biased agonist, partial agonist, etc.). Cell-based assays, such as DiscoveRx PathHunter or those employing the ⌬6-Gqi4-myr protein, are optimized to facilitate high-throughput screening of drug properties. However, as these assays are increasingly utilized for pharmacological assessment, care must be taken in determining whether the differential properties of the ligand detected (such as functional selectivity) can be extrapolated to a physiological context or whether they are artifacts of the cell-based assay utilized in the initial characterization (41). The differences in the results between the two assays used to determine ␤-arrestin2 translocation to the KOR demonstrate how the use of highly modified cellular systems can pose a challenge in determining the properties of a drug. Consistent with the published finding utilizing bioluminescence resonance energy transfer (BRET) (27), the confocal studies presented in this current study reveal that 6Ј-GNTI does not induce any detectable translocation of ␤-arrestin2 to the plasma membrane (Fig. 2B). In contrast, in the KOR-expressing CHO-K1 DiscoveRx PathHunter cell-based assay, 6Ј-GNTI induces low efficacy but high potency recruitment of ␤-arrestin2 to the KOR (Fig.  2A). These differences may be due to intrinsic differences in receptor numbers, assay sensitivity, cell type, time scales, or competition with endogenously expressed, untagged ␤-arres-tin2. Furthermore, the differences could be artifacts of the modified ␤-arrestin2 or KOR constructs used in each assay, encompassing GFP, Renilla luciferase 8, and modified LacZ holoenzyme fusion proteins, which may obscure or distort subtle variations in drug-induced ␤-arrestin2 recruitment. However, 6Ј-GNTI does not induce receptor internalization in either transiently transfected HEK-293 cells (data not shown) or stably transfected CHO-KOR cells, as determined by confocal microscopy, biotinylation, and FACS analysis (Fig. 3) (27).
Moreover, 6Ј-GNTI does not promote KOR-mediated ERK1/2 activation in striatal neurons, a signaling pathway that we show here requires ␤-arrestin2, yet 6Ј-GNTI maintains full efficacy for activating G protein-dependent signaling to Akt in the same neuronal population. Together these observations suggest that the bias for G protein-over ␤-arrestin2-mediated signaling induced by 6Ј-GNTI actions at KOR are preserved in an endogenous setting.
Functionally biased compounds may prove to be useful tools in delineating the relationship between the activation of particular pathways and specific effects at the cellular level in a manner that may also be translated and studied at the behavioral level in animal models, particularly in determining the involvement of KOR interactions with ␤-arrestin2 in the biological effects of KOR ligands. As our data demonstrate, drug screening in more physiologically relevant systems is necessary to truly ascertain the pharmacological profile of a drug. Although the studies in the KOR-expressing cell lines would suggest that KOR-induced activation of ERK1/2 is mediated entirely via pertussis toxin-sensitive G proteins, the studies in striatal neurons demonstrate that KOR signals to ERK1/2 by way of a G proteinindependent, ␤-arrestin2-dependent pathway. This finding may be particularly useful in determining whether newly developed agonists that are biased against ␤-arrestin2 recruitment maintain that bias in the endogenous setting, as the activation of ERK1/2 in neurons can be readily measured. The elucidation of basic drug properties at the isolated receptor in simplified culture systems is still an important part of the drug discovery process. These systems provide the basic understanding of drug/receptor function and capabilities, which can then be modified by the study of different cellular contexts with different complements of receptor-binding proteins and other components of the signaling cascade. However, care must be taken when trying to attribute the signaling responses observed in cell-based assays to the pharmacological profiles obtained in vivo.