Discovery of β-Arrestin–Biased Dopamine D2Ligands
for Probing Signal Transduction Pathways Essential
for Antipsychotic Efficacy
John A. Allena,1, Julianne M. Yostb,1, Vincent Setolaa,1, Xin Chenb, Maria F. Sassanoa, Meng Chenc, Sean Petersond,
Prem N. Yadava, Xi-ping Huanga, Bo Fenge, Niels H. Jensena, Xin Chef, Xu Baif, Stephen V. Fryeb, William C. Wetselc,
Marc G. Carond, Jonathan A. Javitche, Bryan L. Rotha,2, and Jian Jinb,2
aDepartment of Pharmacology and National Institute of Mental Health Psychoactive Drug Screening Program School of Medicine,bCenter for Integrative
Chemical Biology and Drug Discovery, Division of Chemical Biology and Medicinal Chemistry, Eshelman School of Pharmacy, University of North Carolina,
Chapel Hill, NC 27599;cDepartments of Psychiatry and Behavioral Sciences, Cell Biology, and Neurobiology anddDepartments of Cell Biology, Medicine, and
Neurobiology, Duke University Medical Center, Durham, NC 27710;eCenter for Molecular Recognition and Departments of Psychiatry and Pharmacology,
Columbia University, College of Physicians and Surgeons, New York, NY 10032; andfCenter for Combinatorial Chemistry and Drug Discovery, Jilin University,
Changchun, Jilin 130012, China
Edited by Leslie Lars Iversen, University of Oxford, Oxford, United Kingdom, and approved September 29, 2011 (received for review March 26, 2011)
Elucidating the key signal transduction pathways essential for both
antipsychotic efficacy and side-effect profiles is essential for de-
veloping safer and more effective therapies. Recent work has
highlighted noncanonical modes of dopamine D2receptor (D2R) sig-
naling via β-arrestins as being important for the therapeutic actions
of both antipsychotic and antimanic agents. We thus sought to cre-
ate unique D2R agonists that display signaling bias via β-arrestin–
ergic signaling. Through a robust diversity-oriented modification
of the scaffold represented by aripiprazole (1), we discovered
UNC9975 (2), UNC0006 (3), and UNC9994 (4) as unprecedented
β-arrestin–biased D2R ligands. These compounds also represent un-
precedented β-arrestin–biased ligands for a Gi-coupled G protein–
coupled receptor (GPCR). Significantly, UNC9975, UNC0006, and
UNC9994 are simultaneously antagonists of Gi-regulated cAMP pro-
inducing motoric side effects in inbred C57BL/6 mice in vivo. Genetic
deletionofβ-arrestin-2simultaneously attenuatedthe antipsychotic
actions of UNC9975 and transformed it into a typical antipsychotic
drug with a high propensity to induce catalepsy. Similarly, the
antipsychotic-like activity displayed by UNC9994, an extremely
β-arrestin–biased D2R agonist, in wild-type mice was completely
abolished in β-arrestin-2 knockout mice. Taken together, our results
suggest that β-arrestin signaling and recruitment can be simulta-
neously a significant contributor to antipsychotic efficacy and pro-
tective against motoric side effects. These functionally selective,
β-arrestin–biased D2R ligands represent valuable chemical probes
for further investigations of D2R signaling in health and disease.
functional selectivity|ligand bias
nonical pathways involving heterotrimeric large G proteins,
other signaling proteins including, most prominently, β-arrestins
(1–4). The process by which GPCR ligands differentially modulate
canonical and noncanonical signal transduction pathways is a
phenomenon known as “functional selectivity” (5, 6). Such func-
tionally selective ligands preferentially engage either canonical or
noncanonical GPCR pathways (7, 8). Clearly, the discovery of
useful for elucidating the key signal transduction pathways essen-
tial for both the therapeutic actions and the side effects of drugs
(6). Understanding which signaling pathways contribute to anti-
psychotic efficacy and side effects, for instance, will in turn enable
the design of better antipsychotic drug candidates and, ultimately,
lead to safer and more effective therapies for patients. However,
only a small number of functionally selective GPCR ligands have
been reported to date (5–9). In addition to the paucity of such
protein–coupled receptors (GPCRs) signal not only via ca-
ligands, very little purposeful attention has been devoted to cre-
ating and annotating ligands with distinct patterns of functional
selectivity. In particular, although β-arrestin–biased ligands of Gq
and Gs-coupled GPCRs are known (9, 10), β-arrestin–biased
GPCR ligands that selectively activate β-arrestin signaling path-
ways over Gi-coupled pathways have not been reported.
Aripiprazole (OPC-14597, 1), an FDA-approved atypical anti-
psychotic drug, was one of the first functionally selective D2re-
ceptor (D2R or D2) ligands identified (7, 11, 12). Although
aripiprazole wasinitially described as a partial D2R agonist,on the
basis of assays performed in whole animals and isolated tissues
(13–15), it was later demonstrated that aripiprazole could behave
as a full agonist, a partial agonist, or an antagonist at D2R
depending upon the signaling readout and cell type interrogated
(7, 11, 16, 17). However, structure–functional–selectivity rela-
tionships (SFSR) of the aripiprazole scaffold have not been
studied and only modest structure activity relationships (SAR)
have been reported (15, 18). In this paper, we report our discovery
that UNC9975 (2), UNC0006 (3), and UNC9994 (4), analogs
of aripiprazole, represent unprecedented β-arrestin–biased D2R
ligands; these compounds are also unique β-arrestin–biased
ligands for Gi-coupled GPCRs. Here, we describe the design,
synthesis, and in vitro and in vivo activities of these unique
β-arrestin–biased chemical probes. We demonstrate, using wild-
type and β-arrestin-2 knockout mice, that the atypical antipsy-
chotic-like profileoftheseβ-arrestin–biasedD2R agonistsrequires
β-arrestin-2. The development of these unique, functionally se-
lective, β-arrestin–biased D2R ligands provides the biomedical
community with valuable chemical tools for probing signaling
pathways essential for antipsychotic efficacy and side effects.
Discovery of β-Arrestin–Biased Dopamine D2 Ligands UNC9975,
UNC0006, and UNC9994. Because aripiprazole (1) has been shown
to be a functionally selective D2R ligand (7, 11) and because the
SFSRs of the aripiprazole scaffold have not been studied, we
intensely explored four regions of the aripiprazole template to
Author contributions: J.A.A., V.S., S.V.F., W.C.W., M.G.C., J.A.J., B.L.R., and J.J. designed
research; J.A.A., J.M.Y., V.S., X. Chen, M.F.S., M.C., S.P., P.N.Y., X.-p.H., B.F., N.H.J., X. Che,
and X.B. performed research; J.A.A., J.M.Y., V.S., X. Chen, M.F.S., M.C., S.P., P.N.Y., X.-p.H.,
B.F., N.H.J., X. Che, X.B., B.L.R., and J.J. analyzed data; and J.A.A., B.L.R., and J.J. wrote the
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
1J.A.A., J.M.Y., and V.S. contributed equally to this work.
2To whom correspondence may be addressed. E-mail: email@example.com or bryan_
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
www.pnas.org/cgi/doi/10.1073/pnas.1104807108PNAS Early Edition
| 1 of 6
identify D2R ligands with distinct functional selectivity patterns
(Fig. 1). We investigated various substituents (mono- and dis-
ubstitution) on the left-hand side phenyl ring, several cyclic
amino groups, and various linkers in the middle region, along
with a variety of bicyclic aromatic moieties at the right-hand side
of the scaffold. Our intention was to modulate both the con-
formational preferences and substituent binding effects while
retaining the core pharmacophore. In total, we prepared >150
unique compounds using the synthetic procedures described in
SI Methods. These compounds were subsequently evaluated in
dopamine D2receptor binding, Gi-regulated cAMP production,
and G protein–independent β-arrestin-2 translocation assays.
Although most of these compounds (15 representative com-
pounds are summarized in Table S1) strongly activated both
Gi-regulated cAMP accumulation and G protein–independent
β-arrestin-2 recruitment in a manner similar to aripiprazole, we
discovered UNC9975 (2), UNC0006 (3), and UNC9994 (4) as
unprecedented β-arrestin–biased D2R ligands, which were simul-
taneously partial agonists of β-arrestin-2 translocation and antag-
onists of Gi-regulated cAMP production (see below).
UNC9975, UNC0006, and UNC9994 Do Not Activate D2-Mediated Gi-
Regulated Inhibition of cAMP Production. UNC9975, UNC0006,
and UNC9994 were initially evaluated in the D2-mediated cAMP
accumulation assay, which measures inhibition of isoproterenol-
stimulated cAMP production via the Gi-coupled signaling path-
way (19). UNC9975, UNC0006, and UNC9994 did not activate
this Gi-mediated signaling pathway, in stark contrast to aripipra-
zole, which was a potent partial agonist (EC50= 38 nM, Emax=
51%) (Fig. 2A). As expected, quinpirole (20) was a potent full
agonist (Fig. 2A) whereas common typical and atypical antipsy-
chotic drugs were antagonists in this assay. When tested for ag-
onist activity at D2 receptors to decrease basal cAMP in the
absence of isoproterenol, none of the compounds exhibited ac-
tivity. In addition, HEK293T cells devoid of D2receptors were
assayed in parallel and displayed no inhibition of isoproterenol-
stimulated cAMP, either by aripiprazole or by UNC9975 (Fig.
S1), demonstrating that the effect observed in D2 receptor-
expressing cells is solely due to signaling through the D2receptors.
UNC9975, UNC0006, and UNC9994 Induce D2-Mediated β-Arrestin-2
Translocation as Partial Agonists. To assess the effects of these
compounds on recruiting β-arrestin-2 to D2receptors, we used
a D2-mediated β-arrestin-2 translocation Tango assay that is
highly sensitive to β-arrestin-2 recruitment (21). In this assay,
UNC9975, UNC0006, and UNC9994 were potent (EC50< 10
nM) partial agonists for β-arrestin-2 recruitment to D2receptors,
similar to aripiprazole (Fig. 2B). Whereas UNC9975 (Emax=
43%) and UNC0006 (Emax= 47%) were less efficacious than
aripiprazole (Emax= 73%), UNC9994 (Emax= 91%) was more
efficacious than aripiprazole and approached the activity of the
full agonist quinpirole. Haloperidol (22), a typical antipsychotic
(along with all other typical and atypical antipsychotic drugs
UNC0006, and UNC9994 via exploring multiple regions of the aripiprazole
Discovery of unprecedented β-arrestin–biased D2R ligands UNC9975,
β-arrestin–biased dopamine D2partial agonists. (A) Activity of UNC9975,
UNC0006, UNC9994, aripiprazole, and quinpirole in the D2-mediated Gi-
coupled isoproterenol-stimulated cAMP production assay using HEK293T
cells expressing the dopamine D2 receptor and GloSensor-22F. UNC9975,
UNC0006, and UNC9994 did not activate this Gi-mediated signaling pathway
whereas aripiprazole (EC50= 38 nM, pEC50= 7.4 ± 0.1, Emax= 51 ± 5%) was
a partial agonist and quinpirole (EC50= 3.2 nM, pEC50= 8.49 ± 0.07, Emax=
100 ± 3%) was a full agonist. Data are representative of at least two in-
dependent experiments. (B) Activity of UNC9975, UNC0006, UNC9994, ari-
piprazole, quinpirole, and haloperidol in the D2-mediated β-arrestin-2
translocation Tango assay using HTLA cells transfected with a D2V2-TCS-tTA
construct. UNC9975 (EC50= 1.1 nM, pEC50= 8.95 ± 0.03, Emax= 43 ± 0.5%),
UNC0006 (EC50= 1.2 nM, pEC50= 8.91 ± 0.03, Emax= 47 ± 1%), UNC9994
(EC50= 6.1 nM, pEC50= 8.22 ± 0.09, Emax= 91 ± 3%), and aripiprazole (EC50=
2.4 nM, pEC50= 8.62 ± 0.03, Emax= 73 ± 1%) were partial agonists whereas
haloperidol (antagonist control) had no agonist activity. Quinpirole (EC50=
2.0 nM, pEC50= 8.70 ± 0.05, Emax= 100 ± 2%) was used as a positive control.
Data are representative of at least two independent experiments. (C) Ac-
tivity of UNC9975, UNC0006, UNC9994, aripiprazole, and quinpirole in the
D2-mediated β-arrestin-2 translocation DiscoveRx assay with 20 h stimula-
tion. UNC9975 (EC50= 5.7 nM, pEC50= 8.24 ± 0.20, Emax= 19 ± 1%), UNC0006
(EC50= 3.2 nM, pEC50= 8.49 ± 0.15, Emax= 25 ± 1%), UNC9994 (EC50= 448
nM, pEC50= 6.35 ± 0.07, Emax= 64 ± 2%), and aripiprazole (EC50= 3.4 nM,
pEC50= 8.47 ± 0.08, Emax= 51 ± 1%) were partial agonists. Quinpirole (EC50=
56 nM, pEC50= 7.25 ± 0.04, Emax= 100 ± 2%) was used as a positive control.
Data are representative of at least two independent experiments. (D) Ac-
tivity of UNC9975, UNC0006, UNC9994, aripiprazole, and quinpirole in the
D2-mediated BRET-based β-arrestin-2 recruitment assay using HEK293 cells
expressing GRK2. UNC9975 (EC50= 6.0 nM, pEC50= 8.22 ± 0.49, Emax= 20 ±
3%), UNC0006 (EC50= 17 nM, pEC50= 7.77 ± 0.38, Emax= 25 ± 4%), UNC9994
(EC50> 1,000 nM, Emax> 50%), and aripiprazole (EC50= 145 nM, pEC50= 6.84
± 0.18, Emax= 47 ± 4%) were all partial agonists that promote β-arrestin
recruitment to D2receptors. Quinpirole (EC50= 6.7 nM, pEC50= 8.17 ± 0.15,
Emax= 100 ± 5%) was used as a positive control. Data are representative of
at least three independent experiments.
UNC9975, UNC0006, and UNC9994 are functionally selective,
2 of 6
| www.pnas.org/cgi/doi/10.1073/pnas.1104807108Allen et al.
tested) did not activate D2-mediated β-arrestin-2 translocation.
Because HTLA cells (a HEK293-derived cell line stably ex-
pressing a tTA-dependent luciferase reporter gene and a β-
arrestin-2-TEV fusion protein) transfected with a D2V2-TCS-
tTA construct (21) were used in this assay, we tested these
compounds in parallel in HTLA cells transfected with a V2-TCS-
tTA construct (21) to investigate the possibility that the observed
effect was due to compound acting via the V2tail or some other
nonreceptor-mediated arrestin pathway. As expected, UNC9975,
UNC0006, aripiprazole, and quinpirole (a positive control for
D2) were inactive in V2receptor-expressing cells whereas argi-
nine vasopressin (a positive control for V2) was a full agonist
(Fig. S2A). These results demonstrate that the effect of
UNC9975, UNC0006, UNC9994, and aripiprazole observed in
D2V2receptor-expressing cells is due to the compounds acting
via the D2receptor, not at the V2tail.
To confirm this finding from the Tango assay, which has
downstream amplification of signaling, we tested these com-
pounds in an orthologous assay: the D2-mediated β-arrestin-2
recruitment DiscoveRx approach, which uses a β-galactosidase
fragment complementation-based technology to monitor the
interaction of β-arrestin with GPCRs (23). As shown in Fig. 2C
and Fig. S2B, UNC9975 (EC50= 5.7 nM), UNC0006 (EC50=
3.2 nM), and aripiprazole (EC50= 3.4 nM) potently activated
D2-mediated β-arrestin-2 translocation. Although UNC9994
(EC50= 448 nM) was less potent, this compound (Emax= 64%)
had a higher efficacy than aripiprazole (Emax= 51%), UNC9975
(Emax= 19%), and UNC0006 (Emax= 25%) (Fig. 2C). Quin-
pirole (EC50= 56 nM, Emax= 100%) was used as a positive
control in this assay. These findings are consistent with the
results obtained in the D2β-arrestin-2 Tango assay.
Aripiprazole has been previously reported to be either an
antagonist (17) or a partial agonist (24) for β-arrestin-2 recruit-
ment using the bioluminescence resonance energy transfer
(BRET)-based assay. Therefore, we also assessed the effects of
UNC9975, UNC0006, and UNC9994 on D2-mediated β-arrestin-2
recruitment using this BRET-based approach (17). In HEK293T
cells wherein G protein–coupled receptor kinase 2 (GRK2)
along with the D2-dopamine receptor was expressed, aripipra-
zole, UNC9975, UNC0006, UNC9994, and aripiprazole all dis-
played partial agonist activity for D2-mediated β-arrestin-2
recruitment (Fig. 2D). Although UNC9994 was less potent than
aripiprazole, UNC9975, and UNC0006, UNC9994 had a higher
efficacy than aripiprazole, UNC9975, and UNC0006 (Fig. 2D).
Notably, compound activity to recruit β-arrestin-2 was sensitive
to the expression of GRK2 in this BRET-based assay, as all
compounds required GRK2 coexpression to induce detectable
BRET signals in response to β-arrestin-2 recruitment (Fig. 2D
and Fig. S2C). Therefore, using three orthogonal assay plat-
forms, we confirmed that UNC9975 and UNC0006 were potent
partial agonists for β-arrestin-2 recruitment with slightly lower
efficacy than aripiprazole whereas UNC9994 was a higher effi-
cacy and lower potency partial agonist for β-arrestin-2 trans-
location compared with aripiprazole.
We next examined the effects of UNC9975 and UNC0006 on
inducing D2receptor endocytosis (internalization) by assessing
the cell surface fluorescence of D2receptors using flow cytom-
etry (25) in HEK293 cells coexpressing GRK2 and β-arrestin-2.
D2 receptor internalization in response to treatment with
UNC9975, UNC0006, and aripiprazole (1 μM for 1 h) did not
reach statistical significance (Fig. S3). No internalization was
observed in cells treated with haloperidol (as a negative control).
As expected, quinpirole (a positive control) significantly in-
creased receptor internalization.
To assess the effects of UNC9975 and UNC0006 on β-arrestin–
mediated signaling, we evaluated these two compounds in the
extracellular signal-regulated kinase (ERK) phosphorylation
(p-ERK) reporter assay, which is capable of detecting extended
β-arrestin–mediated ERK phosphorylation [following a relatively
long incubation (4 h)] (10). In the HEK293T cells transfected
with D2receptors, UNC9975 (EC50= 2.2 nM, Emax= 32%) and
UNC0006 (EC50= 3.2 nM, Emax= 33%) were potent partial
agonists (Fig. S2A). The potencies and efficacies of UNC9975
and UNC0006 are similar to those of aripiprazole (EC50= 1.8
nM, Emax= 39%) in this assay. Importantly, coexpression of
β-arrestin-2 and GRK2 significantly enhanced the efficacies of
UNC9975, UNC0006, and aripiprazole in this signaling assay
(Fig. S4B). In addition, UNC9975 (Emax= 13%) and UNC0006
(Emax= 11%) were inactive whereas aripiprazole was a potent
partial agonist (EC50= 6.3 nM, Emax= 49%) in the D2dopa-
mine receptor-mediated p-ERK immunofluorescence assay (Fig.
S4C), which measures the rapid phase of G protein–mediated
ERK phosphorylation (5 min incubation) (26). These results are
consistent with the cAMP biosensor assay results (see above) and
further indicate that UNC9975 and UNC0006 lack appreciable
agonism via canonical Gi-dependent pathways.
Taken together, these multiple functional activity profiling
studies demonstrate that UNC9975, UNC0006, and UNC9994
are β-arrestin–biased D2R ligands that selectively activate β-
arrestin recruitment and signaling and are simultaneously in-
active at Gi-mediated signal transduction pathways.
Selectivity of UNC9975, UNC0006, and UNC9994 and PK Parameters of
UNC9975. Consistent with their potencies in D2functional assays,
UNC9975 and UNC0006 displayed high affinities (Ki< 10 nM)
similar to aripiprazole in the D2antagonist radioligand compe-
tition binding assay whereas UNC9994 displayed a lower binding
affinity (Ki= 79 nM) to D2R than UNC9975, UNC0006, and
aripiprazole (Table 1). Although UNC9975, UNC0006, and
UNC9994 also had high affinity for the D3-dopamine receptor,
they displayed low affinities for other dopamine receptors (i.e.,
D1, D4, and D5; Table 1). At serotonin [as known as 5-hydroxy-
tryptamine (5-HT or 5HT)] receptors, UNC9975, UNC0006, and
UNC9994 displayed moderate to high binding affinities (Ki=
0.6–500 nM) for 5HT2A, 5HT2B, 5HT2C, and 5HT1A, but were
significantly less potent in functional assays [Ca2+mobilization
fluorometric imaging plate reader (FLIPR) or cAMP biosensor];
UNC9975, UNC0006, and UNC9994 were antagonists at 5HT2A
and 5HT2B and agonists at 5HT2C and 5HT1A. In addition,
UNC9975, UNC0006, and UNC9994 had relatively high affini-
ties to H1-histamine receptor (Ki< 10 nM) but were less potent
antagonists in H1functional assays. In general, UNC9975 and
UNC0006 displayed a similar GPCR selectivity profile to aripi-
prazole. UNC9994 was less potent against most of these GPCRs
compared with UNC9975, UNC0006, and aripiprazole.
UNC9975, UNC0006, and UNC9994 at select GPCRs
Radioligand binding and functional activities of
Binding affinity or potency, nM*
Aripiprazole UNC9975UNC0006 UNC9994
*Ki, IC50, pA2, or EC50values are the average of at least two duplicate experi-
ments with SD values that are threefold less than the average.
Allen et al.PNAS Early Edition
| 3 of 6
In mouse pharmacokinetic (PK) studies, both UNC9975 and
aripiprazole displayed high exposure levels in brain and excellent
CNS penetration (Table S2). Although the brain exposure level
of UNC9975 was about threefold lower, UNC9975 had a longer
half-life in brain and a higher brain/plasma ratio over 24 h
compared with aripiprazole. The excellent in vivo PK parameters
of UNC9975 make it a suitable tool for in vivo pharmacody-
UNC9975 and UNC9994 Exhibit Antipsychotic Activity in Vivo That Is
Attenuated in β-Arrestin-2 Knockout Mice. To correlate the in vitro
functional selectivity profiles with potential in vivo therapeutic
effects, we evaluated the inhibition of psychostimulant or psy-
of which are well-established pharmacological mouse models for
assessingpotential antipsychoticactivity (27,28).We firstevaluated
the ability of UNC9975 to inhibit D-amphetamine–induced hyper-
locomotion as detailed previously (27) in inbred C57BL/6 mice. As
exhibited an efficacy and potency similar to aripiprazole (ED50=
0.38 mg/kg vs. 0.36 mg/kg) in this in vivo model (Fig. 3B).
We next evaluated the effects of UNC9975 on psychotomi-
metic-induced hyperlocomotion in response to the NMDA re-
ceptor antagonist phencyclidine (PCP) (28). Because β-arrestin-2
knockout mice have a reduced locomotor response to amphet-
amine (3), we used PCP to induce locomotion in these animals. To
assess the potential involvement of β-arrestin in the effects of
UNC9975, β-arrestin-2 knockout mice and wild-type littermate
controls (29) were administered increasing doses of UNC9975
followed by PCP administration. UNC9975 potently and dose
dependently inhibited PCP-induced hyperlocomotion in wild-
type mice (Fig. 3C); however, this antipsychotic-like activity of
UNC9975 was significantly attenuated in β-arrestin-2 knockout
mice (Fig. 3D). Significantly, the in vivo potency of UNC9975
was greatly decreased in β-arrestin-2 knockout mice (β-arrestin-2
knockout ED50= 0.75 mg/kg vs. wild-type ED50= 0.26 mg/kg)
(Fig. 3E). In striking contrast, aripiprazole displayed no significant
differences in potency or efficacy in wild-type (ED50= 0.13 mg/
kg) and β-arrestin-2 knockout mice (ED50= 0.13 mg/kg) (Fig. 3F).
To confirm these findings, we next evaluated the effects of
UNC9994 on PCP-induced hyperlocomotion in wild-type and
β-arrestin-2 knockout mice. UNC9994 (2 mg/kg, i.p.) markedly
inhibited PCP-induced hyperlocomotion in wild-type mice (Fig.
4A). Importantly, this significant antipsychotic-like activity of
UNC9994 was completely abolished in β-arrestin-2 knockout
mice (Fig. 4B).
Because UNC9975 has 5HT2Aantagonist activity (Table 1) and
5HT2Aantagonism can be a significant contributor to the efficacy
of antipsychotic drugs (30, 31), we evaluated SR46349B (a potent
and selective 5HT2A antagonist) (32) and clozapine (a classic
atypical antipsychotic with potent antagonist activity at 5HT2Aand
other aminergic GPCRs) (33) in the PCP-induced hyperlocomo-
tion model in wild-type and β-arrestin-2 KO mice. We found that
SR46349B and clozapine were equally effective at suppressing
PCP-induced locomotion in β-arrestin-2 KO mice vs. wild-type mice
(Fig. S5), indicating that antagonism of 5HT2Areceptors is similar
in both genotypes. These observations suggest that the attenuation
in antipsychotic drug-like activity of UNC9975 in β-arrestin-2 KO
mice is likely not due to activity at 5-HT2Areceptors.
Taken together, our results strongly suggest that the antipsy-
chotic drug-like activity of UNC9975 and UNC9994 requires
β-arrestin and this is likely mediated through the dopamine D2
receptors in mice.
D-amphetamine–induced hyperlocomotion. UNC9975
UNC9975 and UNC0006 Induce Catalepsy in β-Arrestin-2 Knockout
Mice but Not in Wild-Type Mice. To compare potential extrapyra-
midal side effects of UNC9975, UNC0006, and aripiprazole, we
evaluated these compounds along with haloperidol (as a positive
control) in a standard drug-induced catalepsy model (34),
using wild-type and β-arrestin-2 knockout mice. As shown in
Fig. 5 A and B, UNC9975, UNC0006, or aripiprazole (5.0 mg/kg)
failed to significantly induce catalepsy in wild-type mice at either
30 or 60 min after treatment, whereas haloperidol (2.0 mg/kg)
induced significant catalepsy at both time points. In notable
contrast to the results in wild-type animals, the β-arrestin–biased
dopamine D2ligands UNC9975 and UNC0006 significantly in-
duced catalepsy in β-arrestin-2 knockout mice 60 min after
treatment (Fig. 5B). By comparison, aripiprazole, which activates
both Gi-coupled and β-arrestin–mediated pathways, did not in-
duce catalepsy in β-arrestin-2 knockout mice or in wild-type mice
(Fig. 5B). Collectively, these results suggest that β-arrestin re-
cruitment and signaling are protective against motoric side effects.
Through a combined medicinal chemistry and comprehensive
in vitro and in vivo pharmacological profiling approach, we
designed, synthesized, and characterized unique D2β-arrestin–
hyperlocomotion studies that is attenuated in β-arrestin-2 knockout mice.
(A) Locomotor responses in inbred C57BL/6 mice are shown as 5-min binned
intervals to vehicle or different doses (i.p.) of UNC9975 followed 30 min later
by 3 mg/kg D-amphetamine (AMPH, i.p.). (B) Bar graph of distance traveled
after AMPH administration (30- to 70-min time interval). C57BL/6 mice were
given vehicle or different doses of UNC9975 or aripiprazole 30 min before
AMPH treatment. n = 8 animals/group. *P < 0.05, vs. vehicle + 3 mg/kg AMPH
group. (C and D) Locomotor activities shown as 5-min binned intervals of
wild-type (WT) or β-arrestin-2 knockout (β-ARR2 KO) littermate mice given
vehicle or different doses of UNC9975 followed 30 min later with 6 mg/kg
phencyclidine (PCP, i.p.). (E) Bar graph of distance traveled by WT and
β-ARR2 KO mice after PCP administration (30- to 70-min time interval) as
shown in C and D. n = 14 WT and β-ARR2 KO pairs/group. *P < 0.05, vs.
vehicle + 6 mg/kg PCP group. (F) Bar graph of distance traveled by WT and
β-ARR2 KO mice after aripiprazole injection followed by PCP treatment (30-
to 70-min time interval). n = 8 littermate WT and β-ARR2 KO pairs/group.
*P < 0.05, vs. vehicle + 6 mg/kg PCP group.
UNC9975 exhibits potent antipsychotic-like activity in mouse
4 of 6
| www.pnas.org/cgi/doi/10.1073/pnas.1104807108 Allen et al.
biased agonists and demonstrated their atypical antipsychotic
drug-like activities in vivo. These chemical probes represent
unprecedented functionally selective β-arrestin–biased dopamine
D2R ligands that exhibit antipsychotic activity in vivo. This study
thus represents a successful proof-of-concept for how functionally
selective GPCR ligands can be discovered and validated. Addi-
tionally,our findingsshow that onthebasisoftheβ-arrestinbiasof
these compounds, β-arrestin emerges as an important contributor
to both antipsychotic drug efficacy and antipsychotic side effects.
Our robust diversity-oriented synthetic approach capitalized on
the aripiprazole scaffold to generate, initially, hundreds of ana-
logs. Among them, UNC9975, UNC0006, and UNC9994 were
discovered as unique, β-arrestin–biased functionally selective D2
ligands. These compounds are partial agonists that induce D2
receptor-mediated β-arrestin recruitment and signaling and si-
multaneously are inactive at Gi-dependent signaling. Further, it
should be acknowledged that whereas UNC9975, UNC0006, and
UNC9994 promoted the recruitment of β-arrestin to D2recep-
tors, these studies were performed in heterologous cells express-
ing β-arrestin-2 and GRK2. Whether these compounds stimulate
β-arrestin recruitment and signaling in native D2 receptor-ex-
pressing neurons is the subject of ongoing investigations. Inter-
estingly, although aripiprazole, UNC0006, and UNC9975 are all
agonists for β-arrestin recruitment and signaling, in contrast to
quinpirole, none of these ligands promoted robust internalization
of D2receptors, a trafficking phenomenon commonly involving
β-arrestins (35). This inability of aripiprazole to induce significant
receptor internalization is consistent with previous findings (7),
suggesting that a common property of aripiprazole-based ligands
may be the inability to promote robust D2receptor internali-
zation. This lack of internalization may bode well for the eventual
translation of such β-arrestin–biased ligands to the clinic as drugs
that induce internalization would ultimately induce tachyphylaxis
and receptor down-regulation (36).
More significantly, the in vivo activity of UNC9975 indicated
a strict requirement for β-arrestin-2 for both full antipsychotic
activity and protection against motoric side effects. Similar to
aripiprazole, this β-arrestin–biased ligand shows a potent ability
to suppress both D-amphetamine and phencyclidine-induced
hyperlocomotion in mice, indicating that the compound pos-
sesses antipsychotic drug-like activities in vivo. Significantly, the
antipsychotic drug-like activities of UNC9975 were attenuated in
β-arrestin-2 knockout mice, indicating that β-arrestin-2 is re-
quired in vivo for full activity. Interestingly, previous studies have
determined that the locomotor responses induced by the non-
selective direct (apomorphine) and indirect (amphetamine)
agonists are attenuated in β-arrestin-2 knockout mice (3), Im-
portantly, the mechanisms by which β-arrestin influences these
locomotor responses in vivo and to what extent this β-arrestin-2
knockout effect is specific to D2 receptors are currently un-
known. Here we observe that in β-arrestin-2 knockout mice the
unique β-arrestin–biased ligands UNC9975 and UNC9994, but
not the other atypical drugs including aripiprazole, clozapine,
and SR46349B, show an attenuated ability to inhibit PCP-in-
duced locomotion. These findings demonstrate that β-arrestin-2
is required for the full antipsychotic drug-like activity of
UNC9975 and UNC9994. However, further mechanistic studies
will be required to determine how D2-mediated β-arrestin re-
cruitment and signaling by UNC9975 and UNC9994 are essential
for these compounds’ antipsychotic activity in vivo.
With the exception of aripiprazole, all Food and Drug Admin-
istration-approved typical and atypical antipsychotic medications
the common property of antagonizing D2-mediated G protein–
dependent and –independent signaling (17). Indeed, typical and
atypical antipsychotic drugs are antagonists at Gi-mediated signal-
ing and arrestin-ergic pathways. It is currently unknown what, ifany
activity, any of these compounds have at potential Go-mediated
signaling. This antagonism at both signaling pathways is thought to
underlie the therapeutic benefit to prevent psychotic symptoms but
can also cause serious extrapyramidal side effects including cata-
lepsy and other motor dyskenesias (37). Although aripiprazole,
UNC0006,andUNC9975donotinduce catalepsy inwild-type mice
at doses at which antipsychotic drug-like therapeutic activities are
measured, UNC0006 and UNC9975 resemble haloperidol in in-
ducing catalepsy in β-arrestin-2 knockout mice. This propensity to
cause catalepsy in the absence of β-arrestin-2 suggests that
UNC0006 and UNC9975 signal through β-arrestin-2 in vivo and
that this signaling may protect against motoric side effects due to
antagonism at D2R-mediated Gipathways.
These findings have obvious implications for the development
of therapeutic approaches for treating schizophrenia and related
disorders. Atypical antipsychotic drugs, which differ from older
medications (e.g., haloperidol and chlorpromazine) by virtue of
their reduced propensity to induce motoric side effects, are among
the most widely prescribed medications. Many pharmacologic
strategies that aim to target a multiplicity of non-D2receptor
molecular targets (e.g., 5-HT2Ainverse agonists, 5-HT2Cagonists,
mGluR2/3agonists, NK-3 antagonists, sigma antagonists, D1- or
D4-selective antagonists, and so forth) have led to a growing
atypical antipsychotic drugs (37–39). Our results suggest that
atypical antipsychotic drugs with a unique mechanism of action
may arise from β-arrestin–biased D2ligands. Although we do not
know whether such compounds will have therapeutic benefits over
existing medications, it is tempting to speculate that they may
demonstrate special efficacies in psychotic disorders with a mood
dysfunction given the prominent role of arrestin-ergic signaling in
the action of lithium and related compounds (40, 41).
hyperlocomotion studies that is completely abolished in β-arrestin-2 knock-
out mice. (A and B) Locomotor activities shown as 5-min binned intervals for
wild-type (WT) or β-arrestin-2 knockout (β-ARR2 KO) littermate mice given
vehicle or 2.0 mg/kg UNC9994 (i.p.) followed 30 min later with 6 mg/kg
phencyclidine (PCP, i.p.). n = 10–13 WT and β-ARR2 KO pairs/group.
UNC9994 exhibits potent antipsychotic-like activity in mouse
mice but not in wild-type littermates. (A and B) Wild-type (WT) and β-
arrestin-2 knockout (β-ARR2 KO) littermate mice were administered (i.p.)
vehicle, 5.0 mg/kg UNC9975, 5.0 mg/kg UNC0006, 5.0 mg/kg aripiprazole, or
2.0 mg/kg haloperidol. Catalepsy was assessed 30 and 60 min after drug
injection using the inclined screen test where latency to move was scored.
n = 8 WT and β-ARR2 KO pairs/group. *P < 0.05, vs. vehicle controls.
UNC9975 and UNC0006 induce catalepsy in β-arrestin-2 knockout
Allen et al. PNAS Early Edition
| 5 of 6
Methods Download full-text
A complete description of chemical synthesis for all compounds and protocols
for all biological experiments is detailed in SI Methods.1H and13C NMR
spectra of UNC9975 (Fig. S6), UNC0006 (Fig. S7), and UNC9994 (Fig. S8) are
provided in SI Methods.
D2β-Arrestin Recruitment Assay. Recruitment of β-arrestin to agonist-stimu-
lated D2Lreceptors was performed using a previously described “Tango”-type
assay (21). Briefly, HTLA cells stably expressing β-arrestin-TEV protease and
a tetracycline transactivator-driven luciferase were plated into 15-cm dishes in
DMEM containing 10% FBS. Cells were transfected (calcium phosphate) with
clear-bottom, 384-well plates (Greiner; 10,000 cells/well, 50 μL/well) in DMEM
containing 1% dialyzed FBS. The following day, thecells were challenged with
10 μL/well of reference agonist (6 μM) or D2test ligand (6 μM) ± reference
agonist prepared in HBSS, 20 mM Hepes (pH 7.4), and 18% DMSO (final ligand
concentrations are 1 μM, final DMSO concentration is 3%). After 18 h, the
medium was removed and replaced with 1× BriteGlo reagent (Promega), and
luminescence per well was read using a TriLux plate reader (1 s/well). Data
werenormalizedtovehicle(0%)and quinpirole (100%)controlsandregressed
using the sigmoidal dose-response function built into GraphPad Prism 4.0.
In Vivo Studies in Mice. All experiments were approved by the Institutional
and Duke University. C57BL/6J wild-type and β-arrestin-2 knockout mice were
housed under standard conditions: 12-h light/dark cycle with food and water
provided ad libitum. Adult, age-matched male and female wild-type and
β-arrestin-2 knockout drug-naive mice were used for all behavioral testing.
Locomotor activity was assessed under standardized environmental con-
ditions in 21 × 21-cm Plexiglas chambers with photobeams spaced at 2.5 cm
(AccuScan Instruments) as previously described (42). Mice were injected (i.p.)
with vehicle (0.9% saline/0.2% acetic acid), aripiprazole (0.1, 0.25, 0.50, or 2.0
mg/kg), or UNC9975 (0.25, 0.50, or 2.0 mg/kg) and placed into the open field.
For the studies with SR46349B and clozapine, mice were injected (i.p.) with
vehicle (0.9% saline/50 mM tartaric acid), SR46349B (1.0 mg/kg), or clozapine
(1.0 mg/kg) and placed into the open field. Thirty minutes later D-amphet-
amine (3mg/kg)orphencyclidine (6.0 mg/kg) was administered and mice were
immediately returned to the open field. Activity was monitored throughout
this entire period. Horizontal activity was measured as the total distance
traveled in centimeters. The means ± SEMs of the locomotor responses were
analyzed using Graphpad Prism 5.0. To estimate the half-maximal inhibitory
concentration (ED50), dose responses of total locomotor activity during the 90-
min period after D-amphetamine or phencyclidine administration were plot-
ted and best-fit decay curves were determined using a nonlinear regression
one-phase decay equation. Cumulative locomotor responses were analyzed
using a one-way ANOVA followed by Newman–Keuls multiple comparison
tests with Graphpad Prism 5.0. Statistical significance was set at P < 0.05.
ACKNOWLEDGMENTS. We thank Dr. J. Martin Herold for critical reading of
synthetic procedures and characterization data and Dr. Gilad Barnea (Brown
University) for supplying cells and cDNA for the Tango assay. We thank
the National Institutes of Health (Grants U19MH082441S1, U19MH082441,
R01MH61887, R01MH73853, R01DA022413, and R01MH054137) and the
Lieber Center for Schizophrenia Research and Treatment for financial
support and the National Institute of Mental Health Psychoactive Drug
Screen Program for selectivity assay support.
1. Luttrell LM, et al. (1999) Beta-arrestin-dependent formation of beta2 adrenergic
receptor-Src protein kinase complexes. Science 283:655–661.
2. Lefkowitz RJ, Shenoy SK (2005) Transduction of receptor signals by beta-arrestins.
3. Beaulieu JM, et al. (2005) An Akt/beta-arrestin 2/PP2A signaling complex mediates
dopaminergic neurotransmission and behavior. Cell 122:261–273.
4. Abbas A, Roth BL (2008) Arresting serotonin. Proc Natl Acad Sci USA 105:831–832.
5. Urban JD, et al. (2007) Functional selectivity and classical concepts of quantitative
pharmacology. J Pharmacol Exp Ther 320:1–13.
6. Mailman RB (2007) GPCR functional selectivity has therapeutic impact. Trends Phar-
macol Sci 28:390–396.
7. Urban JD, Vargas GA, von Zastrow M, Mailman RB (2007) Aripiprazole has func-
tionally selective actions at dopamine D2 receptor-mediated signaling pathways.
8. Kilts JD, et al. (2002) Functional selectivity of dopamine receptor agonists. II. Actions
of dihydrexidine in D2L receptor-transfected MN9D cells and pituitary lactotrophs.
J Pharmacol Exp Ther 301:1179–1189.
9. Violin JD, Lefkowitz RJ (2007) Beta-arrestin-biased ligands at seven-transmembrane
receptors. Trends Pharmacol Sci 28:416–422.
10. Gesty-Palmer D, et al. (2009) A beta-arrestin-biased agonist of the parathyroid hor-
mone receptor (PTH1R) promotes bone formation independent of G protein activa-
tion. Sci Transl Med 1:1ra1.
11. Lawler CP, et al. (1999) Interactions of the novel antipsychotic aripiprazole (OPC-
14597) with dopamine and serotonin receptor subtypes. Neuropsychopharmacology
12. Mailman RB, Murthy V (2010) Third generation antipsychotic drugs: Partial agonism
or receptor functional selectivity? Curr Pharm Des 16:488–501.
13. Kikuchi T, et al. (1995) 7-(4-[4-(2,3-Dichlorophenyl)-1-piperazinyl]butyloxy)-3,4-dihy-
dro-2(1H)-quinolinone (OPC-14597), a new putative antipsychotic drug with both
presynaptic dopamine autoreceptor agonistic activity and postsynaptic D2 receptor
antagonistic activity. J Pharmacol Exp Ther 274:329–336.
14. Inoue T, Domae M, Yamada K, Furukawa T (1996) Effects of the novel antipsychotic
agent 7-(4-[4-(2,3-dichlorophenyl)-1-piperazinyl]butyloxy)-3,4-dihydro -2(1H)-quinoli-
none (OPC-14597) on prolactin release from the rat anterior pituitary gland. J Phar-
macol Exp Ther 277:137–143.
15. Oshiro Y, et al. (1998) Novel antipsychotic agents with dopamine autoreceptor ago-
nist properties: Synthesis and pharmacology of 7-[4-(4-phenyl-1-piperazinyl)butoxy]-
3,4-dihydro-2(1H)-quinolinone derivatives. J Med Chem 41:658–667.
16. Shapiro DA, et al. (2003) Aripiprazole, a novel atypical antipsychotic drug with
a unique and robust pharmacology. Neuropsychopharmacology 28:1400–1411.
17. Masri B, et al. (2008) Antagonism of dopamine D2 receptor/beta-arrestin 2 interaction
is a common property of clinically effective antipsychotics. Proc Natl Acad Sci USA 105:
18. Johnson DS, et al. (2011) Discovery of PF-00217830: Aryl piperazine napthyridinones
as D2 partial agonists for schizophrenia and bipolar disorder. Bioorg Med Chem Lett
19. Kimple AJ, et al. (2009) Structural determinants of G-protein alpha subunit selectivity
by regulator of G-protein signaling 2 (RGS2). J Biol Chem 284:19402–19411.
20. Tsuruta K, et al. (1981) Evidence that LY-141865 specifically stimulates the D-2 do-
pamine receptor. Nature 292:463–465.
21. Barnea G, et al. (2008) The genetic design of signaling cascades to record receptor
activation. Proc Natl Acad Sci USA 105:64–69.
22. Granger B, Albu S (2005) The haloperidol story. Ann Clin Psychiatry 17:137–140.
23. McGuinness D, et al. (2009) Characterizing cannabinoid CB2 receptor ligands using
DiscoveRx PathHunter beta-arrestin assay. J Biomol Screen 14:49–58.
24. Klewe IV, et al. (2008) Recruitment of beta-arrestin2 to the dopamine D2 receptor:
Insights into anti-psychotic and anti-parkinsonian drug receptor signaling. Neuro-
25. Guo N, et al. (2010) Impact of D2 receptor internalization on binding affinity of
neuroimaging radiotracers. Neuropsychopharmacology 35:806–817.
26. Strachan RT, Sciaky N, Cronan MR, Kroeze WK, Roth BL (2010) Genetic deletion of p90
ribosomal S6 kinase 2 alters patterns of 5-hydroxytryptamine 2A serotonin receptor
functional selectivity. Mol Pharmacol 77:327–338.
27. Schlumberger C, et al. (2009) Comparison of the mGlu(5) receptor positive allosteric
modulator ADX47273 and the mGlu(2/3) receptor agonist LY354740 in tests for
antipsychotic-like activity. Eur J Pharmacol 623:73–83.
28. Nordquist RE, et al. (2008) Effects of aripiprazole/OPC-14597 on motor activity,
pharmacological models of psychosis, and brain activity in rats. Neuropharmacology
29. Bohn LM, et al. (1999) Enhanced morphine analgesia in mice lacking beta-arrestin 2.
30. Meltzer HY (1999) The role of serotonin in antipsychotic drug action. Neuro-
psychopharmacology 21(2, Suppl):106S–115S.
31. Yadav PN, et al. (2011) The presynaptic component of the serotonergic system is re-
quired for clozapine’s efficacy. Neuropsychopharmacology 36:638–651.
32. Rinaldi-Carmona M, et al. (1992) Biochemical and pharmacological properties of SR
46349B, a new potent and selective 5-hydroxytryptamine2 receptor antagonist.
J Pharmacol Exp Ther 262:759–768.
33. Coward DM (1992) General pharmacology of clozapine. Br J Psychiatry Suppl, 5–11.
34. Chipkin RE, et al. (1988) Pharmacological profile of SCH39166: A dopamine D1 se-
lective benzonaphthazepine with potential antipsychotic activity. J Pharmacol Exp
35. Ferguson SS (2001) Evolving concepts in G protein-coupled receptor endocytosis: The
role in receptor desensitization and signaling. Pharmacol Rev 53:1–24.
36. Allen JA, Roth BL (2011) Strategies to discover unexpected targets for drugs active at
G protein-coupled receptors. Annu Rev Pharmacol Toxicol 51:117–144.
37. Roth BL, Sheffler DJ, Kroeze WK (2004) Magic shotguns versus magic bullets: Selectively
38. Conn PJ, Roth BL (2008) Opportunities and challenges of psychiatric drug discovery:
Roles for scientists in academic, industry, and government settings. Neuro-
39. Gray JA, Roth BL (2007) The pipeline and future of drug development in schizo-
phrenia. Mol Psychiatry 12:904–922.
40. Beaulieu JM, Gainetdinov RR, Caron MG (2009) Akt/GSK3 signaling in the action of
psychotropic drugs. Annu Rev Pharmacol Toxicol 49:327–347.
41. Beaulieu JM, et al. (2008) A beta-arrestin 2 signaling complex mediates lithium action
on behavior. Cell 132:125–136.
42. Abbas AI, et al. (2009) PSD-95 is essential for hallucinogen and atypical antipsychotic
drug actions at serotonin receptors. J Neurosci 29:7124–7136.
6 of 6
| www.pnas.org/cgi/doi/10.1073/pnas.1104807108Allen et al.