Functional Selectivity of Dopamine D1 Receptor Signaling: Retrospect and Prospect

Research progress on dopamine D1 receptors indicates that signaling no longer is limited to G protein-dependent cyclic adenosine monophosphate phosphorylation but also includes G protein-independent β-arrestin-related mitogen-activated protein kinase activation, regulation of ion channels, phospholipase C activation, and possibly more. This review summarizes recent studies revealing the complexity of D1 signaling and its clinical implications, and suggests functional selectivity as a promising strategy for drug discovery to magnify the merit of D1 signaling. Functional selectivity/biased receptor signaling has become a major research front because of its potential to improve therapeutics through precise targeting. Retrospective pharmacological review indicated that many D1 ligands have some degree of mild functional selectivity, and novel compounds with extreme bias at D1 signaling were reported recently. Behavioral and neurophysiological studies inspired new methods to investigate functional selectivity and gave insight into the biased signaling of several drugs. Results from recent clinical trials also supported D1 functional selectivity signaling as a promising strategy for discovery and development of better therapeutics.


A Brief Introduction of Functional Selectivity
The term functional selectivity [1] was first introduced in 1994 and soon thereafter other reports referred to this phenomenon alternately as "agonist trafficking of signaling", "differential engagement", or "biased agonism". The concept of functional selectivity essentially means that a ligand may have different actions at two or more signaling pathways mediated by the same receptor. Ligand actions may range from full agonism at one pathway and antagonism at another. Before functional selectivity was introduced, a ligand was defined as either an agonist or antagonist for one canonical signaling pathway mediated by the target receptor. This simplified notion had to be abandoned once it became clear that multiple signaling pathways could be involved by the activation of a receptor. This pharmacological insight contributed significantly to the development of the concept of functional selectivity. Meanwhile, functional selectivity also benefited from the progress on structural biology where the discovery of heterodimers and some detailed crystal structures inspired understanding of additional signaling properties of a receptor. The concept of functional selectivity has evolved. Initially, functional selectivity was studied using a "paired model", such that a newly discovered receptor signaling pathway was compared with the traditional canonical pathway. Now, the "network model" is more appreciated whereby all signaling pathways related to a receptor are treated equally and the signaling complex as a whole accounts for the pharmacological property of a ligand. Functional selectivity has become a major pharmacological research front and more relevant as additional signaling pathways were discovered. The full signaling properties of many new and some "old" ligands have been studied or reinvestigated to better understand their engagement of targets in more precise ways and potentially to improve their therapeutic index. To grasp the significance of functional selectivity of dopamine D 1 ligands, it is important first to Figure 1. Dopamine D1 receptor-related signaling. The traditionally canonical G protein coupled cAMP signaling potentially could be subdivided based on G protein subtype [3] and PKA subunit [4]. G protein independent, β-arrestin-related signaling acts through MAP kinase phosphorylation [5], and has cross talk with cAMP signaling [6,7]. Receptor recycling is also regulated by β-arrestin. Regulation of ion channels could be through cAMP [8]. Gαq dependent PLC signaling is controversial [9]. Abbreviations: D1R, dopamine D1 receptor; AC5, adenylate cyclase type 5; PKA, protein kinase A; ERK, extracellular-signal-regulated kinase; GRK, G protein-coupled receptor kinase; DARPP-32, Dopamine and cAMP-related phosphoprotein 32KDa; Rap, a small GTPase; CREB, cAMP response element-binding protein; PLC, phospholipase C. G proteins are a family of proteins made up of subunits Gα, Gβ, and Gγ. D1Rs stimulate adenylate cyclase primarily through Gαs. The Gαs family is comprised of Gαs and Golf, the latter named for its predominant expression in the olfactory system. Studies on Gαolf knock-out mice suggested that Gαolf may play an essential role in D1R-mediated cAMP accumulation [10,11]. Gαolf knock-out mice showed no hyperlocomotor response to the D1 agonist SKF81297, and their striatal D1Rs had decreased affinity for dopamine. The adenylyl cyclase response to dopamine in the caudate/putamen and nucleus accumbens were also decreased dramatically, but prefrontal cortex signaling remained unaffected. Studies on the striatum of Parkinson's disease patients also suggested that Gαolf plays the essential role in D1R-mediated cAMP [12]. The hallmark of Parkinson's disease is decreased dopamine in the striatum. Interestingly, Gαolf in this region is less abundant in Parkinson's patients, whereas Gαs is highly expressed. The combination of dopamine deficiency and Gαolf being sparsely expressed suggested that the coupling of D1Rs to adenylate cyclase is mediated primarily by Gαolf or the switch/shift between Gαs and Gαolf. The traditionally canonical G protein coupled cAMP signaling potentially could be subdivided based on G protein subtype [3] and PKA subunit [4]. G protein independent, β-arrestin-related signaling acts through MAP kinase phosphorylation [5], and has cross talk with cAMP signaling [6,7]. Receptor recycling is also regulated by β-arrestin. Regulation of ion channels could be through cAMP [8]. Gα q dependent PLC signaling is controversial [9]. Abbreviations: D 1 R, dopamine D 1 receptor; AC5, adenylate cyclase type 5; PKA, protein kinase A; ERK, extracellular-signal-regulated kinase; GRK, G protein-coupled receptor kinase; DARPP-32, Dopamine and cAMP-related phosphoprotein 32KDa; Rap, a small GTPase; CREB, cAMP response element-binding protein; PLC, phospholipase C. G proteins are a family of proteins made up of subunits Gα, Gβ, and Gγ. D 1 Rs stimulate adenylate cyclase primarily through Gα s . The Gα s family is comprised of Gα s and G olf , the latter named for its predominant expression in the olfactory system. Studies on Gα olf knock-out mice suggested that Gα olf may play an essential role in D 1 R-mediated cAMP accumulation [10,11]. Gα olf knock-out mice showed no hyperlocomotor response to the D 1 agonist SKF81297, and their striatal D 1 Rs had decreased affinity for dopamine. The adenylyl cyclase response to dopamine in the caudate/putamen and nucleus accumbens were also decreased dramatically, but prefrontal cortex signaling remained unaffected. Studies on the striatum of Parkinson's disease patients also suggested that Gα olf plays the essential role in D 1 R-mediated cAMP [12]. The hallmark of Parkinson's disease is decreased dopamine in the striatum. Interestingly, Gα olf in this region is less abundant in Parkinson's patients, whereas Gα s is highly expressed. The combination of dopamine deficiency and Gα olf being sparsely expressed suggested that the coupling of D 1 Rs to adenylate cyclase is mediated primarily by Gα olf or the switch/shift between Gα s and Gα olf .
The mRNA of other G proteins are also expressed in the striatum. For example, G protein γ 7 subunit mRNA was detected within rat brain areas that aligned with striatumenriched adenylyl cyclase, dopamine receptors, and Gα olf . This suggested that γ 7 may be part of the Gα olf -containing complex that couples dopamine receptors selectively to adenylyl cyclase [13]. Another study used a ribozyme strategy to suppress the expression of the G protein γ 7 subunit in HEK 293 cells stably expressing the human D 1 Rs, and revealed a significant attenuation of D 1 agonist SKF81297-stimulated adenylyl cyclase activity [14]. This evidence supported the potentially important function of the G protein γ 7 subunit on D 1 R-mediated cAMP signaling.
Along with improved understanding of the interaction between D 1 Rs and each G protein subunit, there is ongoing progress regarding functional selectivity of G protein subtype-specific signaling. Although there are currently no D 1 ligands being developed to target specific G protein subunits, such compounds have been reported for the C-C chemokine receptor 5 (CCR5) [3]. Lorenzen et al. studied four chemokine analogs and found that some analogs were super agonists for G q/11 activation, whereas other analogs displayed a signaling bias for G i/o . Their results demonstrated that ligands can elicit G protein subtype-specific signaling bias and cause receptors to couple preferentially to one subtype of G protein signaling over others. This is inspiring for the pharmacology field. A timely study is needed to investigate this interesting phenomenon since it is reasonable to assume that some dopamine ligands may also engage each G protein subtype differently, providing a more targeted action.

β-arrestin-related Signaling and Mitogen-Activated Protein (MAP) Kinase Activation
In addition to G protein-dependent adenylyl cyclase signaling, research has demonstrated the importance of G protein independent β-arrestin-related signaling. Originally, arrestin was thought to only promote receptor desensitization and internalization. It blocks GPCR coupling to G proteins, preventing GPCR's activation (desensitization), or links GPCR to elements of internalization machinery. Starting in the 1970s, Nobel Laureate Robert Lefkowitz and his colleagues made several discoveries related to β-adrenergic receptors, and β-arrestin related signaling is one of their concomitant discoveries. To date, it has been reported widely in several receptor systems, including β-adrenergic [5], opioid [15], cannabinoid [16], angiotensin II type 1 [17], 5-HT2A [18], apelin [19], growth hormone secretagogue [20], sphingosine 1-phosphate [21], and dopamine D 2 -like receptors (D 2 Rs) [22,23]. Regarding D 1 Rs, there are a few reports that suggested the existence of an interaction between D 1 R and β-arrestin and indicated the involvement of MAP kinase phosphorylation ( Figure 1), but many of them did not specify or propose D 1 R-mediated β-arrestin signaling.
Chen et al. used co-immunoprecipitation to show that extracellular signal-regulated kinases 1 and 2 (ERK 1/2 ) formed stable heterotrimeric complexes with the D 1 R and β-arrestin2. In cells transfected with the dominant negative mutant of β-arrestin2, however, the formation of such complexes was inhibited substantially [24]. Similarly, Urs et al. utilized a D 1 R knockout mouse model and demonstrated that formation of the β-arrestin2 and ERK 1/2 complexes was also blunted [25,26]. In contrast, others stimulated striatal or prefrontal cortex D 1 Rs in vivo or in vitro with the D 1 selective agonist SKF38393 and showed ERK 1/2 was phosphorylated [24,27,28]. These studies provided the initial evidence of D 1 R-mediated β-arrestin signaling and its function through ERK 1/2 . In addition to ERK 1/2 , several studies indicated the involvement of other MAP kinases. Zhen et al. showed SKF38393 increased activation of p38 MAP kinase and c-Jun amino-terminal kinase in SK-N-MC neuroblastoma cells that endogenously express D 1 Rs, whereas ERK activity was not affected [29]. This study suggested that D 1 R-related MAP kinase phosphorylation could be cell-type specific.
Another interesting finding is that D 1 R-related MAP kinase phosphorylation may also be dependent of PKA. Studies in parkinsonian mouse models indicated that a PKA substrate, dopamine and cAMP-regulated 32 kDa phosphoprotein (DARPP-32), was critically involved in D 1 R-mediated ERK 1/2 phosphorylation. A mutation on the phosphorylation site of DARPP-32 reduced activation of ERK, whereas sensitization of DARPP-32 led to increased activation of ERK 1/2 [6,30]. The selective PKA inhibitor Rp-cAMP eliminated the activations of ERK 1/2 , p38 MAP kinase, and c-Jun amino-terminal kinase [27,29]. In contrast, cAMP directly activated a guanine-nucleotide-exchange factor that stimulates Rap GTPase and promotes the MAP kinase cascade [7,31]. These studies suggested there may be potential cross talk between D 1 R-mediated cAMP and β-arrestin-related signaling.

Regulation of Ca 2+ , K + , and Na + Channels
Many studies have indicated D 1 R involvement in the regulation of ion channels. Voltage-dependent Ca 2+ channels (L-, N-, and P/Q-type) play critical roles in balancing intracellular Ca 2+ concentrations that are key for neurotransmitter release and synaptic plasticity [32]. Using whole-cell voltage-clamp techniques, Surmeier et al. showed that the application of D 1 agonists reduced N-and P-type Ca 2+ currents, but enhanced L-type currents. The differential regulation of Ca 2+ currents by D 1 agonists helps to explain the diversity of effects that D 1 Rs have on synaptic integration and plasticity [33]. Hernandez-Lopez et al. furthered the study of D 1 R effects on L-type Ca 2+ current. They examined the impact of D 1 agonists at depolarized and hyperpolarized membrane potentials and showed that D 1 R activation either can inhibit or enhance evoked activity, depending on the level of membrane depolarization. Interestingly, the effects on evoked activity at membrane potentials were blocked by the L-type Ca 2+ channel antagonists nicardipine or calciseptine, and were occluded by the agonist BayK8644. These data indicated that the D 1 R-mediated effects occurred through the L-type Ca 2+ channel [34]. For N-type Ca 2+ channels, coimmunoprecipitation showed the existence of a D 1 R-N-type Ca 2+ channel signaling complex in the prefrontal cortex. This complex had a direct receptor-channel interaction. D 1 agonists not only regulated N-type Ca 2+ channel distribution but also inhibited influx Ca 2+ current. Consequently, neuronal transduction was attenuated [35,36]. The cAMP/PKA/DARPP-32 signaling cascade appeared to mediate these effects on Ca 2+ channels, as cyclic AMP analogs mimicked the effects of D 1 agonists [33].
The D 1 agonist SKF81297 or SKF82957 in combination with the D 2 agonist quinpirole increased spike firing of nucleus accumbens neurons via inhibition of a slow A-type K + current. This enhancement was prevented by inhibitors of PKA or G βγ and enabled by intracellular perfusion with G βγ . These data suggested that the underlying mechanism of D 1 R and D 2 R cooperativity in mediating the slow A-type K + current was by activation of specific subtypes of adenylyl cyclases released by G βγ from the G i/o -linked D 2 R in combination with Gα s -linked D 1 R [37]. The D 1 agonists SKF81297 or dihydrexidine induced prolonged membrane depolarization and excitability of fast-spiking interneurons in the prefrontal cortex. Voltage-clamp analyses revealed that this mimicked dopamine-suppressed inward rectifying K + current and can be reduced by the D 1 antagonist SCH23390 [38][39][40][41]. Although the precise mechanism underlying D 1 R-mediated K + current changes has not been fully understood, studies have suggested possible options: the direct interaction of cAMP with K + channels and the involvement of D 1 R-mediated cAMP/PKA signaling. The first reason is that the effect of D 1 R stimulation on K + current can be mimicked by the adenylyl cyclase activator forskolin and the active cAMP analog Sp-cAMP. The second reason is that the inhibition of PKA with either PKI, Rp-cAMP, or the protein phosphatase inhibitor okadaic acid abolished D 1 R modulation [40,41]. D 1 Rs also appear to impact Na + channels. For example, the D 1 agonist SKF38393 reduced the peak Na + current amplitude in rat striatal neurons and subsequently depressed striatal neuron excitability. These effects were reversed by the D 1 antagonist SCH23390 [42,43]. Intracellular loading of PKA mimicked D 1 R-mediated Na + current inhibition, and diffusion of the PKA inhibitor PKI into the cytosol of neurons blocked it, suggesting the involvement of PKA [44]. Schiffmann et al. suggested the critical role of phosphorylated DARPP-32 as part of this pathway since its injection reduced the Na + current amplitude [45]. This line of evidence suggested that D 1 R regulation of ion channels may be a subsequent event of D 1 -mediated G protein-dependent cAMP signaling. On the other hand, Cantrell et al. reported that phosphorylation of Ser573 on the Na + channel α subunit was critical for D 1 R-mediated effects on the Na + current since this site was phos-phorylated by D 1 R activation [46]. Since the structure of the ion channel itself potentially could play a critical role, as shown by the Ser 573 study, it is also reasonable to assume that there is a more "direct" interaction between the D 1 R and ion channel. A timely study is needed to investigate this interesting hypothesis. More importantly, if some dopamine ligands can engage each ion channel differently, they may provide a more targeted action and potentially lead to better therapeutic implications.

Phospholipase C (PLC) Activation
D 1 R-mediated PLC signaling was once proposed as a novel target, but controversies occurred, and it now is considered to be purported. The possibility that D 1 Rs may function through PLC first was reported in a series of studies on adenylate cyclase type 5 (AC5), a dopamine sensitive-adenylate cyclase. Genetic disruption of the AC5 isoform led to loss of adenylyl cyclase activity after administering the D 1 agonist SKF38393, and this was accompanied by a decrease in the expression of Gα s . AC5 null mice also showed parkinsonian-like motor dysfunction. Interestingly, administration of the partial D 1 agonist SKF38393 improved some of the symptoms, suggesting compensation of D 1 signaling outside the Gα s mechanism and beyond adenylate cyclase [47,48]. Gα q -mediated PLC activation and subsequent Ca 2+ elevation as a non-cyclase signaling for D 1 Rs was then proposed to explain D 1 R-mediated motor behaviors of the null mice. SKF38393 increased Gα q protein binding to the D 1 R in the striatum, suggesting the possible role of the Gα q protein in D 1 R-mediated PLC activation [49]. Several studies indicated that SKF38393 activated PLC in brain slices, and this action was inhibited selectively by the D 1 antagonist SCH23390. In addition, dopamine-induced inward Ca 2+ current was mimicked by the administration of SKF38393 and blocked by SCH23390 [50][51][52].
The most supportive evidence for D 1 R-mediated PLC signaling came from studies using the D 1 ligand SKF83959 that has small effects on adenylate cyclase but strong efficacy for PLC activation. Interestingly, it induced contralateral rotations in the unilateral 6-OHDA-lesioned parkinsonian rat model, and the rotations were completely blocked by the D 1 antagonist SCH23390 [53,54]. The involvement of the D 1 R in PLC signaling, however, is still controversial [9] because the behavioral effects of SKF83959 can be explained by several other mechanisms. First, SKF83959 is still a typical partial agonist for adenylate cyclase [55]. Second, non-specific effects on other receptors also could explain the behavioral effects of SKF83959 since it can bind to several GPCRs in micromolar concentrations [56]. Third, it is postulated that D 1 Rs and D 2 Rs form a D 1 /D 2 heterodimer. Heterodimers have been shown to play a role in functional selectivity in several other GPCR systems [57,58], including the D2/neurotensin NTS1 receptor complex [59] and the D2/trace amine-associated receptor 1 heterodimer [60]. It is very tantalizing to think that the D 1 /D 2 heterodimer led to PLC signaling. D 1 Rs and D 2 Rs, however, are seldom co-expressed in striatal neurons [61][62][63][64][65], suggesting that the heterodimer mechanism is likely not a major contributor. Collectively, the evidence suggests that D 1 Rs actually may be independent of PLC activation.

Insight on Functional Selectivity through the Implications of D 1 Signaling
Dopamine receptors are highly expressed in the brain. The densest area is forebrain where the major dopaminergic terminal fields occur including caudate-putamen and nucleus accumbens. The midbrain (i.e., substantia nigra and ventral tegmental area) also has a high density of dopamine receptors. Olfactory, limbic, and brainstem areas have moderate densities of dopamine receptors. The cerebral cortex has a relatively light density of dopamine receptors but those that are there have important functional implications. In general, the density of D 1 Rs (including D 1 and D 5 ) is higher than that of D 2 Rs (D 2 , D 3 and D 4 ), especially in the cortex where there is a significantly higher overall density and different laminar patterning of D 1 R compared to D 2 R. The D 1 R is preferentially distributed in deeper cortex layers and is proportionally more widespread and expressed within local GABAergic interneuron populations. In the basal ganglia where the density of dopamine receptors is the highest, the segregation of D 1 R and D 2 R is more distinct, with <10% overlap. The GABAergic medium spiny projection neurons of the striatum express D 1 R in the direct pathway and D 2 R in the indirect pathway. These two parallel and segregated pathways form the outflow of the basal ganglia to regulate thalamocortical circuitry [66]. Dysfunction of dopamine receptors in these brain areas play causal roles in many neurological disorders. Therefore, targeting D 1 Rs for therapeutic intervention is attractive. In this section, we try to differentiate D 1 signaling by highlighting some reports that in retrospect have contributed to the understanding of functional selectivity.
There has been a long history between D 1 R-mediated cAMP and Parkinson's disease. AC5 is highly concentrated in the striatum. Genetic ablation of the AC5 gene eliminated adenylate cyclase activity stimulated by D 1 agonists in the striatum, and induced parkinsonian-like motor dysfunction. These findings supported the involvement of D 1 R-mediated AC5 activation in the motor symptoms of Parkinson's disease [47,48,67]. AC5-produced striatal cAMP binds to the regulatory subunits of PKA that then phosphorylates various proteins such as DARPP-32 and cAMP response element-binding protein (CREB). Although how this signaling leads to D 1 -mediated behavioral effects is still unclear, these downstream molecules are involved in the regulation of gene expression [4]. These lines of evidence encouraged the development of functionally selective dopamine ligands whose cAMP signaling can be biased to the PKA subunit to provide a more targeted action improved therapeutic index.
Recent studies on D 1 R-mediated β-arrestin have yielded several impressive clinical implications. Urs et al. reported that the D 1 R-dependent, β-arrestin-related ERK signal cascade affected morphine-induced psychomotor activation but not reward [24,25], suggesting a separation of therapeutics (e.g., analgesic) from side effects (e.g., addiction). By analyzing transcriptional signatures in humans and mice, Labonte et al. reported that D 1 R-mediated β-arrestin signaling through ERK may effect sex-specific depression [68]. Several studies on rats or mice with mutations that eliminate β-arrestin recruitment showed less locomotor activity [69], more dyskinesia-like behavior [26], enhanced adiposity [20], and impaired memory reconsolidation [70]. Recently, our team used a pair of D 1 agonists with distinct signaling profiles at β-arrestin recruitment to evaluate rodent behavior in a working memory-related T-maze task [71]. We showed subtle but significant behavioral variation associated with the level of β-arrestin recruitment, suggesting a promising implication of β-arrestin-selective D 1 agonists on cognitive improvement. Interestingly, there were also clinical implications reported from ligands that bias against β-arrestin. Jiang et al. and other groups reported that β-arrestin activation was related to β-amyloid-induced cognitive impairment [72][73][74]. Since β-amyloid is one of the key players in Alzheimer's disease, this finding implies that bias against β-arrestin could be a target for limiting βamyloid-induced cognitive impairment. D 1 agonists that have less β-arrestin activity may be novel therapeutics. Functionally selective D 1 agonists that have less β-arrestin activity may have less side effects in Parkinson's disease because ERK 1/2 activation correlated with levodopa-induced dyskinesia, whereas blocking ERK 1/2 activation significantly decreased it [30]. Indeed, recent phase III Parkinson's disease and phase IIa schizophrenia trials using novel D 1 agonists with almost no β-arrestin activity have shown highly positive clinical indications [75][76][77][78][79][80].
It is encouraging to see D 1 R signaling has several clinical implications, even though some results seemed to contradict each other. In fact, the concept of functional selectivity was born in part to deal with the fact that different types of bias at a signaling pathway lead to benefits or disadvantages depending on different conditions. Essentially, functional selectivity was founded on the idea of precise targeting. In other words, the bias at one receptor's whole signaling complex could be adjusted based on different applications to magnify merit.

Progress on the Structural Biology of Dopamine Receptors
Progress on the structural biology of dopamine receptors has a large impact on the theory of functional selectivity and the discovery of subtype selective ligands, although it is rather slow for D 1 Rs. The crystal structure of D 1 R complexed with a G protein and a noncatechol agonist was reported only recently [81], although predictions of the 3D structure of human D 2 Rs, the binding site, and binding affinities for agonists and antagonists have been around since the 2000s [82]. Over the intervening years, several crystal structures of ligandbound D 2 R, D 3 R, D 4 R, and D 2 R-G i complexes were reported [83][84][85]. By studying selected transmembrane-5 serine mutations, Fowler et al. showed that receptor conformations were involved in D 2L R functional selectivity [86]. Using docking simulations and site-directed mutagenesis, Zhang et al. not only reported the crystal structure of the human angiotensin II type 1 receptor in complex with one of its inverse agonists but also identified specific interactions between the angiotensin II type 1 receptor and different ligands. This provided support for the structural basis of ligand recognition and functional selectivity [87].
The heterodimer theory was also enhanced by structural biology progress. Crystal structures of receptor-ligand complexes allow for rational design of novel molecular probes and drugs. Following this strategy, bivalent ligands were synthesized and investigated for the D 2 R/neurotensin NTS1 receptor heterodimer. The results indicated a strong, predominantly NTS1 receptor-mediated β-arrestin-2 recruitment in D 2 R/NTS1 receptorco-expressing cells [59]. There have been several other studies on the functional selectivity of GPCR heterodimerization [57,58] but because of the sparse co-expression of D 1 Rs and D 2 Rs in the same neuron [61][62][63][64][65], it still is unclear whether the D 1 /D 2 heterodimer plays a critical role in cell signaling or overt behaviors.

Potential of Receptor Localization-Related Functional Selectivity
It is commonly known that dopamine receptors function differently in distinctive brain areas. Although there is not yet many focused studies to specifically investigate this, the possibility is high that functional selectivity is involved as a part of the underlying mechanism. More intriguingly, recent studies have shown that besides the primary effects in the central nervous system, dopamine also acts in the pancreas as a peripheral regulator of metabolism, and this action is functionally biased at D 2 Rs [88]. This suggests that the receptor localization-related functional selectivity may be a widespread phenomenon. GPCRs not only function at the plasma membrane but also at various cellular organelles including endosomes, mitochondria, and Golgi [89]. Although D 1 -related studies are sparse, it will not be a surprise that future studies show D 1 functional selectivity in different organelles. Location bias is an emerging paradigm in GPCR biology and drug discovery. This is particularly true for endosomes regarding D 1 Rs. D 1 Rs start to internalize through endosomes where β-arrestin is recruited. It has been shown that endocytosis promotes rapid D 1 signaling [90]. This could be in line with the initial signal at the plasma membrane, but there are studies showing GPCR signaling via heterotrimeric G proteins from endosomes [91]. Therefore, it will not be unexpected that D 1 has signaling bias toward β-arrestin at endosomes, but more studies are needed for this topic.

Functionally Selective D 1 Ligands: Pharmacological Retrospect
Functional selectivity is currently a hot topic in the drug discovery field. Recently, several novel, functionally selective D 1 ligands were reported [92,93]. They all were biased toward cAMP compared to β-arrestin signaling. Functional selectivity, at the extreme, has a ligand act as both an agonist and antagonist in different cellular functions at a single receptor [1]. An example would be functioning as an agonist at cAMP and an antagonist at β-arrestin (Figure 2). For many dopamine ligands, however, it is more common to activate all signaling pathways but to different degrees. This incomplete/subtle functional selectivity was first reported by our group using a series of full and partial D 1 agonists [71]. Full and partial agonists were defined by activation of the canonical D 1 signaling pathwaythe intrinsic activity at stimulating adenylate cyclase and producing cAMP. It is noteworthy that a retrospective review of some "traditional/classical" D 1 ligands may have revealed that they are functionally selective even if this was not appreciated at the time of publication, as highlighted by our group [71].
at β-arrestin (Figure 2). For many dopamine ligands, however, it is more common to activate all signaling pathways but to different degrees. This incomplete/subtle functional selectivity was first reported by our group using a series of full and partial D1 agonists [71]. Full and partial agonists were defined by activation of the canonical D1 signaling pathway-the intrinsic activity at stimulating adenylate cyclase and producing cAMP. It is noteworthy that a retrospective review of some "traditional/classical" D1 ligands may have revealed that they are functionally selective even if this was not appreciated at the time of publication, as highlighted by our group [71]. Lewis et al. compared the intrinsic activities at adenylate cyclase for a series of structurally dissimilar full and partial D1 agonists and their ability to cause functional desensitization [94]. Surprisingly at the time, they found a dissociation between these two activities. For example, three full D1 agonists (dihydrexidine, SKF82958, A77636) caused homologous desensitization of the D1R in vitro to the same extent as dopamine, whereas two other full agonists (dinapsoline, A68930) and all the partial agonists (SKF38393, pergolide, LSD) caused only partial desensitization that was homologous, but not associated with PKA-induced phosphorylation. Ryman-Rasmussen et al. later tested 13 D1 agonists and showed that internalization efficacy was independent of either agonist affinity or chemotype [95]. For example, four agonists from two chemotypes were able to activate adenylate cyclase fully without inducing internalization. A follow-up study further confirmed that ligand-specific differential effects on receptor recycling involved aspects of D1Rs that are distal from the ligand-binding domain [96]. Since receptor recycling (i.e., desensitization, internalization) is initiated by β-arrestin recruitment, it is logical to conclude that a retrospective review of these data actually provides proof that these compounds have some degree of functional selectivity between canonical cAMP signaling and novel β-arrestinrelated signaling. Moreover, these data are consistent with the hypothesis that functional selectivity reflects subtle ligand-induced conformational changes, as opposed to simple agonist trafficking among discrete receptor active states.

Prospect on Future Studies of Function Selectivity
Functional selectivity was first reported as a pharmacological phenomenon. Later studies using behavioral and neurophysiological methods, and recent clinical trials, have led to improved understanding of functionally selective drugs and provided significant insight on how to further investigate this "so called" pharmacological phenomenon. The Figure 2. Diagram of functional selectivity. Complete functional selectivity, at the extreme, is a ligand acting as both an agonist and an antagonist at different functions at a single receptor. The left panel exemplifies complete bias with a ligand functioning as an agonist (red color) only at Gα olf -mediated cAMP activation. The middle panel exemplifies another complete bias with a ligand functioning as an antagonist (blank) only at β-arrestin-related ERK activation. Incomplete/subtle functional selectivity more commonly is seen in many dopamine ligands. The right panel exemplifies mild bias since all signaling pathways are activated but to different degrees. Each pie represents a signaling pathway; red color indicates pathways are activated but different transparencies reflect they are activated to different levels (e.g., full vs. partial). Note the signaling pathways, such as Gα olf -mediated cAMP activation or β-arrestin related ERK activation, are only examples.
Lewis et al. compared the intrinsic activities at adenylate cyclase for a series of structurally dissimilar full and partial D 1 agonists and their ability to cause functional desensitization [94]. Surprisingly at the time, they found a dissociation between these two activities. For example, three full D 1 agonists (dihydrexidine, SKF82958, A77636) caused homologous desensitization of the D 1 R in vitro to the same extent as dopamine, whereas two other full agonists (dinapsoline, A68930) and all the partial agonists (SKF38393, pergolide, LSD) caused only partial desensitization that was homologous, but not associated with PKA-induced phosphorylation. Ryman-Rasmussen et al. later tested 13 D 1 agonists and showed that internalization efficacy was independent of either agonist affinity or chemotype [95]. For example, four agonists from two chemotypes were able to activate adenylate cyclase fully without inducing internalization. A follow-up study further confirmed that ligand-specific differential effects on receptor recycling involved aspects of D 1 Rs that are distal from the ligand-binding domain [96]. Since receptor recycling (i.e., desensitization, internalization) is initiated by β-arrestin recruitment, it is logical to conclude that a retrospective review of these data actually provides proof that these compounds have some degree of functional selectivity between canonical cAMP signaling and novel β-arrestin-related signaling. Moreover, these data are consistent with the hypothesis that functional selectivity reflects subtle ligand-induced conformational changes, as opposed to simple agonist trafficking among discrete receptor active states.

Prospect on Future Studies of Function Selectivity
Functional selectivity was first reported as a pharmacological phenomenon. Later studies using behavioral and neurophysiological methods, and recent clinical trials, have led to improved understanding of functionally selective drugs and provided significant insight on how to further investigate this "so called" pharmacological phenomenon. The earliest example of functional selectivity was for a series of D 2 ligands that were full agonists at adenylate cyclase but antagonists at other functions [97][98][99]. The translated in vivo behavioral effects of these highly functionally selective compounds indicated they were uncharacteristic of typical D 2 agonists [100,101]. Contemporaneous studies of OPC-9 of 14 14597 (later named aripiprazole) showed that its unusual properties most likely involved functional selectivity at the D 2 Rs and possibly 5-HT 1A receptors [1,102,103], contrary to later views of it functioning as a simple partial agonist. Spurred by studies emerging at the same time in serotonin [104], angiotensin [17], and opioid [105] systems, both basic research and drug discovery have exploded in the past decade with numerous exciting findings on functional selectivity. One timely area relates to the opioid epidemic in which the search for functionally selective opioid receptor ligands indicated superior analgesic action with decreased addictive or other unwanted properties [106][107][108][109]. Recently, our group reported a landmark neurophysiological study on the functional selectivity of D 1 R-mediated cAMP and β-arrestin signaling. Using a pair of D 1 agonists with distinct signaling profiles, we evaluated rodent behavior in a T-maze task and examined how this was associated with neural activities in the prefrontal cortex [71]. We showed significant neurophysiological changes correlated with the level of β-arrestin recruitment. These results indicated the feasibility of using neurophysiological measurements as markers for studying D 1 R functional selectivity. It is encouraging that the field has discovered more ways to investigate functional selectivity-not only by pharmacological means but also behavioral and neurophysiological methods. These interdisciplinary approaches improve the innovation and development of more functionally selective D 1 ligands as better therapeutics.

Summary
The understanding of functional selectivity has evolved over the years in conjunction with advances in improved knowledge of fundamental receptor signaling and complexes. With the discovery of many novel signaling or sub-pathways related to D 1 Rs, studies related to grasping the breadth of D 1 R functional selectivity are expanding. Even though some reports at the time of publishing did not focus on functional selectivity, a retrospective review of their findings indicate they contributed to this field. More importantly, many studies showed positive implications for each unique D 1 signaling pathway, suggesting that functional selectivity could be a promising strategy for drug discovery. Furthermore, retrospective pharmacological review revealed that many D 1 ligands have some degree of mild functional selectivity. Moreover, novel compounds with extreme bias at D 1 signaling were reported recently. Collectively, these data show that the development of precision medicine with the use of functionally selective D 1 ligands is a promising direction to pursue.