Identification and Characterization of ML321: A Novel and Highly Selective D2 Dopamine Receptor Antagonist with Efficacy in Animal Models That Predict Atypical Antipsychotic Activity

Abstract

We have developed and characterized a novel D2R antagonist with exceptional GPCR selectivity - ML321. In functional profiling screens of 168 different GPCRs, ML321 showed little activity beyond potent inhibition of the D2R and to a lesser extent the D3R, demonstrating excellent receptor selectivity. The D2R selectivity of ML321 may be related to the fact that, unlike other monoaminergic ligands, ML321 lacks a positively charged amine group and adopts a unique binding pose within the orthosteric binding site of the D2R. PET imaging studies in non-human primates demonstrated that ML321 penetrates the CNS and occupies the D2R in a dose-dependent manner. Behavioral paradigms in rats demonstrate that ML321 can selectively antagonize a D2R-mediated response (hypothermia) while not affecting a D3R-mediated response (yawning) using the same dose of drug, thus indicating exceptional in vivo selectivity. We also investigated the effects of ML321 in animal models that are predictive of antipsychotic efficacy in humans. We found that ML321 attenuates both amphetamine- and phencyclidine-induced locomotor activity and restored pre-pulse inhibition (PPI) of acoustic startle in a dose-dependent manner. Surprisingly, using doses that were maximally effective in both the locomotor and PPI studies, ML321 was relatively ineffective in promoting catalepsy. Kinetic studies revealed that ML321 exhibits slow-on and fast-off receptor binding rates, similar to those observed with atypical antipsychotics with reduced extrapyramidal side effects. Taken together, these observations suggest that ML321, or a derivative thereof, may exhibit ″atypical″ antipsychotic activity in humans with significantly fewer side effects than observed with the currently FDA-approved D2R antagonists.

Figure 1

ML321 shows high selectivity for the D2R among dopamine and related biogenic amine receptors. (A) Structure of ML321. (B) Radioligand binding competition assays to determine ML321 affinity for dopamine receptors. Radioligand binding assays were performed as described in the Methods. The data are expressed as percentage of the control specific binding and represent mean ± SEM values from three independent experiments each performed in triplicate. Mean ML321 K_i values [95% C.I.] for each receptor were calculated from the IC₅₀ values using the Cheng–Prusoff equation and found to be 57.6 nM [45.7–72.8] for the D2R and 3.9 μM [2.8–5.5] for the D3R (because the curve for the D3R was not complete, the K_i for ML321 should be considered an approximation). The IC₅₀ values for the D1R, D4R, and D5R were > 10 μM. (C) Psychoactive Drug Screening Program (PDSP) results. Radioligand binding assays were performed as described in the Methods. Each bar represents a unique drug target (Table S1). This process identified four receptors for which 10 μM ML321 inhibited the specific binding by more than 50% (defined as the level of significance by the PDSP): D2R (92%) and D3R (59%), 5-HT_2C (64%), and 5-HT₇ (53%) serotonin receptors.

Figure 2

Functional profiling of ML321 against an array of 168 known GPCRs. A single high concentration (10 μM) of ML321 was screened using a β-arrestin recruitment assay in the DiscoverX gpcrMAX assay panel in both antagonist (A) or agonist (B) modes as described in the Methods. Data represent the percent maximum stimulation (agonist mode) observed by a reference agonist for each GPCR or the percent inhibition (antagonist mode) of a response produced by an EC₈₀ concentration of a reference agonist for each GPCR. A complete key to the GPCR array and numerical results are provided in Table S2. Only responses that deviated >20% from the baseline are considered to be significant. In panel (A), these include the following receptors (% activity): D2R_L (98%), D2R_S (98%), D3R (72%), 5-HT_2A (48%), 5-HT_2C (35%), BLT1 (leukotriene B₄) (LTB4R) (37%), sphingosine-1-phosphate 4 (S1PR4) (36%), α_2C-adrenergic (31%); and in panel (B), CB2 cannabinoid (44%).

Figure 3

Curve-shift assays indicate that ML321 behaves in a competitive manner with dopamine at the D2R. D2R-mediated β-arrestin recruitment assays (A) or cAMP inhibition assays (LANCE) (B) were conducted by stimulating the receptor with the indicated concentrations of dopamine with or without various concentrations of ML321 as described in the Methods. For the cAMP assay, the cells were incubated with 10 μM forskolin to stimulate cAMP production. Data are expressed as a percentage of the maximum dopamine response seen in the absence of ML321 (% control) and represent the mean ± SEM of at least three independent experiments each performed in triplicate. Insets show Schild analyses of the data from which mean K_B values [95% C.I.] were derived. (A) ML321 K_B = 103 nM [76.7–127] (n = 3), (B) ML321 K_B = 8.36 nM [3.5–16.5] (n = 3).

Figure 4

ML321 exhibits inverse agonist activity at the D2R. Go BRET activation assays were performed as described in the Methods. Data represent the mean ± SEM of 4–6 independent experiments each performed in octuplicate and are expressed as ΔBRET from the baseline, which represents the constitutive BRET signal seen in the absence of any drug treatment. Dopamine produces a dissociation of the Go heterotrimer and thus a decrease in BRET, whereas due to constitutive activity of the D2R, the tested antagonists exhibit inverse agonist activity through the promotion of heterotrimer formation. The ΔBRET signals for each compound were statistically different from the baseline as determined using Dunnett’s multiple comparison test following a one-way ANOVA (* indicates p < 0.05 and ** indicates p < 0.0001). The ΔBRET signals for the antagonists did not statistically differ from each other (p = 0.841 using one-way ANOVA).

Figure 5

Computationally identified binding pose of ML321 at the D2R. (A) Side view of the D2R model in complex with several docked poses of ML321 (shown in orange sticks). The extracellular and intracellular sides of the receptor are on the top and bottom of the figure, respectively. The location of the ligand binding pocket is indicated by a dotted box. (B) Zoom-in view of the ligand binding pocket bound with the converged binding pose of ML321. The key contact residues are shown in green representations. Note the bulky hydrophobic or aromatic side chains of Ile184^EL2.52, Phe389^6.51, and Phe390^6.52 are tightly and complementarily packed with the dibenzothiazepine moiety of ML321, while the thiophene moiety protrudes into a subpocket formed by Val91^2.61, Leu94^2.64, Trp100^EL1.50, Phe110^3.28, and Cys182^EL2.50.

Figure 6

Binding of ML321 to the D2R is regulated by Na⁺. D2R and D3R radioligand binding assays were performed as described in the Methods using Tris buffer in the absence or presence of 140 mM NaCl. The data are expressed as percentage of the control specific binding and represent mean ± SEM values from three independent experiments each performed in triplicate. Mean ML321 K_i values [95% C.I.] for each receptor were calculated from the IC₅₀ values using the Cheng–Prusoff equation. (A) D2R competition binding curves ± NaCl. In the presence of Na⁺, the mean ML321 K_i = 72.5 nM [56.3–93.3] whereas in the absence of Na⁺ the K_i >10 μM. (B) D3R competition binding curves ± NaCl. In the absence or presence of Na⁺, the ML321 K_i values are >10 μM.

Figure 7

Displacement of the micro-PET tracer [¹¹C]SV-III-130 by ML321 in brains of non-human primates. MicroPET imaging studies were conducted as described in the Methods. Uptake of the radioactive micro-PET tracer [¹¹C]SV-III-130 was monitored continuously for 100 min after injection (i.v.). For the drug treatments, either 1 or 5 mg/kg of ML321 was administered (i.v.) 20 min (arrows) after tracer infusion. Top: coronal brain sections are shown illustrating the uptake of [¹¹C]SV-III-130 into the caudate and putamen (orange areas) under baseline conditions (left) or after the administration of 1 mg/kg (middle) or 5 mg/kg (right) of ML321 (t = 100 min). Bottom: tissue time-activity curves illustrating the uptake of [¹¹C]SV-III-130 into the caudate (left), putamen (middle), or cerebellum (right). The data are normalized to the maximum uptake seen in each brain region. After 20 min (arrows), the animals were treated with either the vehicle or indicated doses of ML321. The representative results from single experiments are shown, which were performed three times with similar results.

Figure 8

Effects of ML321 on D2R-mediated hypothermia or D3R-mediated yawning in rats. Measurements of hypothermia and yawning were performed as described in the Methods. All data are presented as the mean ± SEM. (A) Measurement of hypothermia: change in rats’ body temperature following injection (s.c.) with vehicle (0.0 mg/kg) or 1.0, 3.2, or 10.0 mg/kg ML321 and 1.0 mg/kg sumanirole (SUM). Experimental groups contained 6 animals each except for the 10.0 mg/kg ML321 + vehicle group, which contained 3 animals. For the ML321 + sumanirole groups following a significant effect of ML321 dose in one-way ANOVA [F(3,20) = 14.69, p < 0.001], pairwise comparisons to vehicle were made post hoc using Dunnett’s tests (two-tailed): ***p < 0.001, significantly different from 0.0 mg/kg ML321 (vehicle) + 1.0 mg/kg sumanirole. For the 10 mg/kg ML321 + vehicle group, the data are presented for reference but were not statistically analyzed given the small number of animals tested. (B) Measurement of yawning: number of yawns made in 60 min following injection (s.c.) with vehicle (0.0 mg/kg) or 3.2 or 10.0 mg/kg ML321 and 0.1 mg/kg pramipexole (PRAM). Experimental groups contained 8 animals except for the 10.0 mg/kg ML321 + vehicle group, which contained 3 animals. For the ML321 + pramipexole groups, one-way ANOVA for dose was not significant. [F(2,21) = 0.03, p = 0.96]. For the 10 mg/kg ML321 + vehicle group, the data are presented for reference but were not statistically analyzed given the small number of animals tested. (C) To obtain a time course of pramipexole-induced yawning, the 60 min observation session in panel (B) was divided into 4 blocks of 15 min each, and the data were reanalyzed as the mean yawns in each time-block. Two-way ANOVA detected a significant effect of block [F(3,84) = 16.54, p < 0.001]; however, the mean yawns/blocks were not affected by the ML321 dose (non-significant main effect and interaction).

Figure 9

ML321 displays potent antipsychotic-like activity in a hyperlocomotion study. C57BL/6J mice were injected (i.p.) with either the vehicle (Veh) or the indicated doses of ML321 (ML) and placed in an open field with assessment of locomotor activity as described in Methods. After 30 min, the mice were removed and injected (i.p.) with either Veh, 3 mg/kg amphetamine (AMPH), or 6 mg/kg phencyclidine (PCP) and immediately returned to the open field for a further 90 min. Locomotor activities (distance traveled) are shown as 5 min binned intervals (A–C) or as cumulative locomotion (D, E). For the AMPH experiment (B, D), a RMANOVA for the baseline (0–30 min) identified a significant effect of time [F(5,285) = 25.168, p < 0.001], while an analysis of the post-injection interval (31–120 min) revealed the time [F(17,969) = 20.532, p < 0.001], treatment [F(5,57) = 18.231, p < 0.001], and time by treatment interaction [F(85,969) = 3.860, p < 0.001] were significant. For the PCP study (C, E), a RMANOVA for the baseline (0–30 min) observed a significant effect of time [F(5,335) = 28.707, p < 0.001], whereas an analysis of the stimulated interval (31–120 min) reported the time [F(17,1139) = 102.322, p < 0.001], treatment [F(6,67) = 29.915, p < 0.001], and the time by treatment interaction [F(102,1139) = 12.401, p < 0.001] to be significant. (D, E) One-way ANOVA found that baseline (0–30 min) locomotor activities were not significantly different between the Veh–Veh and 5 mg/kg ML–Veh groups. However, following administration of the psychostimulants (31–120 min), one-way ANOVAs demonstrated significant stimulated effects for the AMPH [F(5,62) = 18.231, p < 0.001] and PCP [F(6,73) = 26.915, p < 0.001] experiments. All data are presented as means ± SEMs. N = 11 for the 0.25 ML-PCP group, N = 13 for the Veh–Veh control, and N = 10 for all other groups. *p < 0.05, Veh-AMPH or Veh–PCP vs the other indicated groups; ^#p < 0.05, 0.25 mg/kg ML–PCP vs the Veh–Veh, 5 mg/kg ML–Veh, and the 0.5–5 mg/kg ML–PCP groups; ^{^}p < 0.05, 0.5 mg/kg ML–AMPH vs the 5 mg/kg ML–AMPH group; ⁺p < 0.05, 1 mg/kg ML–AMPH vs the Veh–Veh and 5 mg/kg ML–Veh groups.

Figure 10

ML321 reverses psychostimulant-induced impairments in pre-pulse inhibition of the acoustic startle. C57BL/6J mice were administered (i.p.) vehicle (Veh) or 0.25, 0.5, 1.0, or 5 mg/kg doses of ML321 (ML) in their home cages. Ten minutes later, animals were given the Veh, 3 mg/kg amphetamine (AMPH) (A), or 6 mg/kg phencyclidine (PCP) (B) and acclimated to a 65 dB white-noise background for 10 min in the PPI apparatus. After 5 min, mice were presented with combinations of startle (120 dB), pre-pulse-pulse (4, 8, and 12 dB over the 65 dB background followed by 120 dB), and null trials over 25 min (see Methods). Activity was recorded during all trials. The data are presented as % PPI = [1 – (pre-pulse trials/startle-only trials)] × 100. A RMANOVA for the AMPH study noted significant effects of PPI [F(2,122) = 147.032, p < 0.001], treatment [F(6,61) = 6.120, p < 0.001], and the PPI by treatment interaction [F(12,122) = 2.366, p = 0.009]. A RMANOVA for the PCP experiment determined that the effects of PPI [F(2,122) = 176.947, p < 0.001], treatment [F(6,61) = 8.393, p < 0.001], and the PPI by treatment interaction [F(12,122) = 1.883, p = 0.043] were significant. The data are presented as means ± SEMs. N = 9 mice for the vehicle–vehicle, vehicle–AMPH, and vehicle–PCP groups; N = 10 mice for all other groups. *p < 0.05, for the AMPH study: Veh–AMPH vs the Veh–Veh, 5 mg/kg ML–Veh, and the 5 mg/kg ML–AMPH groups; or *p < 0.05, for the PCP experiment: Veh–PCP vs the Veh–Veh, 5 mg/kg ML–Veh, and the 0.5 to 5 mg/kg ML–PCP groups; ^#p < 0.05, for the AMPH investigation: 0.25 mg/kg ML–AMPH vs the 5 mg/kg ML–Veh and the 5 mg/kg 5 ML–AMPH groups; or for the PCP study: ^#p < 0.05, 0.25 mg/kg ML–PCP vs the Veh–Veh and the 5 mg/kg ML–Veh groups.

Figure 11

ML321 has low cataleptic potential in C57BL/6J mice compared to haloperidol in the horizontal bar test. Baseline responses were recorded and then separate groups of C57BL/6J mice were injected with either vehicle (Veh) or the indicated doses of haloperidol (HAL) or ML321 (ML). Mice were tested 60 min later for catalepsy. The latency for a mouse to remove its paws from the bar was recorded as an index of catalepsy (60 second maximum time). A RMANOVA observed the main effects of time [F(1,72) = 85.874, p < 0.001] and treatment [F(7,72) =35.542, p < 0.001] as well as the time by treatment interaction [F(7,72) = 37.267, p < 0.001] to be significant. The data are presented as means ± SEMs. N = 10 mice/group. *p < 0.05, baseline vs 60 min; ^{^}p < 0.05, 1 mg/kg HAL vs 0.01 and 0.1 mg/kg HAL, and all doses of ML; ⁺p < 0.05, 10 mg/kg HAL vs 0.01 and 0.1 mg/kg HAL, and all doses of ML.