The dopamine D2 receptor can directly recruit and activate GRK2 without G protein activation

The dopamine D2 receptor (D2R) is a G protein–coupled receptor (GPCR) that is critical for many central nervous system functions. The D2R carries out these functions by signaling through two transducers: G proteins and β-arrestins (βarrs). Selectively engaging either the G protein or βarr pathway may be a way to improve drugs targeting GPCRs. The current model of GPCR signal transduction posits a chain of events where G protein activation ultimately leads to βarr recruitment. GPCR kinases (GRKs), which are regulated by G proteins and whose kinase action facilitates βarr recruitment, bridge these pathways. Therefore, βarr recruitment appears to be intimately tied to G protein activation via GRKs. Here we sought to understand how GRK2 action at the D2R would be disrupted when G protein activation is eliminated and the effect of this on βarr recruitment. We used two recently developed biased D2R mutants that can preferentially interact either with G proteins or βarrs as well as a βarr-biased D2R ligand, UNC9994. With these functionally selective tools, we investigated the mechanism whereby the βarr-preferring D2R achieves βarr pathway activation in the complete absence of G protein activation. We describe how direct, G protein–independent recruitment of GRK2 drives interactions at the βarr-preferring D2R and also contributes to βarr recruitment at the WT D2R. Additionally, we found an additive interaction between the βarr-preferring D2R mutant and UNC9994. These results reveal that the D2R can directly recruit GRK2 without G protein activation and that this mechanism may have relevance to achieving βarr-biased signaling.

G protein-coupled receptors (GPCRs) 2 signal through two main transducers: G proteins and ␤-arrestins (␤arrs) (1)(2)(3). G proteins regulate second messenger cascades such as cAMP and calcium that translate extracellular signals into cellular responses (4). Following prolonged or high concentrations of agonist stimulation, activated G proteins and their second messengers cause the recruitment and activation of GPCR kinases (GRKs) (5,6). GRKs phosphorylate serine and threonine residues on the intracellular domains of GPCRs, which promotes the recruitment of ␤arrs. The binding of ␤arr mediates receptor desensitization and endocytosis by sterically blocking further G protein activation and by serving as an adaptor for the AP-2 and clathrin components of the endocytic machinery (7,8). Additionally, the scaffolding function of ␤arr enables it to bring together various other signaling components to mediate G protein-independent signaling (9 -11).
Recently there has been a surge of interest in targeting GPCRs with biased or functionally selective ligands. These types of ligands favor the engagement of one signaling pathway over the other (12,13). Functionally selective drugs are predicted to have increased specificity of action by only engaging the therapeutically relevant pathway while avoiding activation of potential side effect-inducing pathways (14 -16). For the dopamine D2 receptor (D2R), selective targeting of the D2R/ ␤arr signaling pathway may improve drugs used to treat schizophrenia, Parkinson's disease, or other disorders (15,17,18).
However, the question of how fully activated D2R signal transduction cascades might be segregated remains unanswered. This is because maximal binding of ␤arr occurs at GRK-phosphorylated and agonist-occupied GPCRs (19,20). The five nonvisual GRKs that act on GPCRs rely on distinct plasma membrane recruitment mechanisms (21). For GRK2/3, which are the major GRKs that interact with D2Rs, plasma membrane recruitment is driven by an interaction between activated G␤␥ subunits and a C-terminal domain in GRK2/3 (5,22). Thus, the binding of ␤arr appears to be intimately linked to G protein-mediated recruitment of GRKs.
However, there is precedent for direct recruitment and activation of GRK2 by GPCRs. A notable study by Beautrait et al. (23) showed both G protein-and GPCR-mediated components of GRK2 recruitment and mapped the residues on the N terminus of GRK2 involved in mediating recruitment and/or phosphorylation. Indeed, recent structural work at the ␤2AR and GRK5 demonstrates that GRK5 activation occurs via a rearrangement of GRK5's RH/catalytic domain, which facilitates interaction between GRK's catalytic domain and GPCR's intracellular loops (24). These interdomain interactions in GRKs are also necessary for holding them in inactive conformations (25). Therefore, interaction between the agonist-occupied GPCR and GRKs is necessary for activating GRKs and can facilitate their plasma membrane recruitment. Here we sought to understand how GRK2 action at the D2R would be disrupted when G protein activation is eliminated and the effect of this on ␤arr recruitment.
We have previously generated two D2R mutants that display a high degree of signaling bias between the G protein and ␤arr pathways (26). The G protein preferring D2R ( [Gprot] D2R) retains the ability to activate G proteins while having markedly reduced ␤arr recruitment, whereas the ␤arr preferring D2R ( [␤arr] D2R) loses engagement of G proteins while retaining ␤arr recruitment.
Additionally, our laboratory contributed to the development and characterization of a ␤arr-biased D2R ligand, UNC9994 (27). UNC9994 is based on the chemical scaffold of aripiprazole and has essentially no activity at the G protein pathway but retains partial agonism at the ␤arr2 pathway. The degree of UNC9994's agonist/antagonist activity at the ␤arr pathway depends on GRK2 expression levels (28). When GRK2 expression is low, UNC9994 behaves more as a ␤arr-biased antagonist, and when GRK2 expression is high, it gains agonist activity at the ␤arr pathway (28).
In this study, we combine these tools to investigate how loss of G protein activation impacts GRK2 engagement by the D2R and the resulting effect on ␤arr recruitment. We find that direct, G protein-independent recruitment of GRK2 by the D2R plays a significant role and is a means whereby the D2R can selectively promote D2R/␤arr interactions. The implications of these findings were further explored using the GRK2-dependent ligand UNC9994, where we found an additive interaction between it and the ␤arr-preferring D2R mutant. The elucidation of this and other mechanisms of achieving ␤arr bias should inform future efforts to design functionally selective ligands at the D2R and other GPCRs.

No appreciable G protein coupling of the [␤arr] D2R
We first wanted to understand the mechanism whereby the [␤arr] D2R achieves ␤arr recruitment without apparent G protein activation. Originally, the biased D2R mutants were characterized at the G protein pathway using the GloSensor assay, which measures downstream cAMP as a proxy for G protein activation (26). Therefore, we first tested whether potentially low levels of G protein activation by the [␤arr] D2R could be responsible for GRK2 recruitment to this receptor. We employed a recently described TGF␣ shedding assay to monitor G protein activation over time (29). Importantly, this assay is sensitive enough to detect even basal GPCR activity (29). A schematic depicting how this assay works is presented in Fig.  1A. We have adapted this assay to work with the fluorescent alkaline phosphatase substrate 4-methylumbelliferyl phosphate. For more details, see "Experimental procedures" and Ref. 29.
When performed with a G␣ i1/2 chimeric G q protein, we observe the predicted shedding of the alkaline phosphatase into the culture medium with the [WT] D2R and [Gprot] D2R (Fig. 1B). In contrast, there is no activity of the [␤arr] D2R at G␣ i1/2 (Fig. 1B), which matches the inactive [D80A] D2R mutant (30). When performed with a G␣ i3 chimeric G q protein (Fig. 1C), we observe similar results. A negative control chimeric G q protein that lacks the six C-terminal residues also showed no activity with any D2R (Fig. 1D). When combined with our previous results using the GloSensor assay (26), we conclude that the [␤arr] D2R completely lacks functional G protein coupling. These data support the hypothesis that the [␤arr] D2R is capable of inducing ␤arr recruitment independent of G protein activation.

Pertussis toxin reduces ␤arr2 recruitment to [WT] D2R and [Gprot] D2R but not to [␤arr] D2R
We next used bioluminescent resonance energy transfer (BRET) assays to measure the interaction between the D2R and ␤arr2 or GRK2 (31, 32) ( Fig. 2A). To remove the contribution of G␤␥ signaling (and its recruitment of GRK2/3) to the recruitment of ␤arr2, we pretreated the cells overnight with 200 ng/ml pertussis toxin (PTX). PTX ADP ribosylates G␣ i/o proteins, which completely inhibits G␣ i/o activation and functional coupling (33)(34)(35). As shown with a ␤arr2 BRET assay in Fig. 2, pretreatment with PTX reduced the efficacy and potency of ␤arr2 recruitment to the [WT] D2R (Fig. 2B) and the efficacy of ␤arr2 recruitment to the [Gprot] D2R (Fig. 2D). In contrast, the recruitment of ␤arr2 to the [␤arr] D2R was not affected by PTX (Fig. 2C). This result suggests that G protein-mediated translocation of GRK2 is not necessary for the recruitment of ␤arr2 to the [␤arr] D2R. The level of [WT] D2R's ␤arr recruitment has nearly the same efficacy as [␤arr] D2R's. Because the G protein inhibition caused by PTX did not eliminate ␤arr2 recruitment to the [WT] D2R or the [Gprot] D2R, an alternative mechanism may enable the D2R to recruit ␤arr2 in the absence of G protein signaling, whereby GRK2 is directly recruited to the D2R.

Pharmacological and genetic inhibition of GRK2/3 kinase activity diminishes ␤arr2 recruitment to WT and biased D2R mutants
We next tested how inhibition of GRK2 kinase activity affected recruitment of ␤arr2 to the D2Rs. We and others have shown previously that recruitment of ␤arr2 to the [WT] D2R depends on GRK2/3 kinase activity at both endogenous and heterologously expressed D2Rs (28,36). We focused on GRK2 because it is expressed in the dopamine system (37), and D2R has been reported previously not to undergo significant GRK4/ 5/6-mediated phosphorylation in vitro (38).
We first tested this dependence pharmacologically by pretreating the cells with a selective GRK2/3 inhibitor, Cmpd101. Cmpd101 works by binding to the kinase domain active site of GRK2/3 with an IC 50 of ϳ35 nM (39) or 290 nM (40), depending on the assay used. Importantly, the IC 50 for the related kinase GRK5 is Ͼ125 M when tested with an in vitro kinase assay (39). Because we do not know the plasma membrane permeability of Cmpd101, we pretreated the cells for 60 min with 10 M Cmpd101 for these assays to ensure that the IC 50 for GRK2 was reached. As shown in Fig. 3, A-C, pretreatment with Cmpd101 markedly reduces ␤arr2 recruitment to the WT and biased D2R mutants. This result indicates that GRK2/3 kinase activity is necessary for the maximal recruitment of ␤arr2 to the D2R.
We then confirmed this phenomenon genetically by using a dominant negative K220R mutant GRK2 that lacks kinase activity but can still bind G␤␥ subunits and be recruited to the plasma membrane (41). Again, we observed that disruption of normal GRK2 kinase activity significantly reduced ␤arr2 recruitment to the D2R (Fig. 3, D-F). Together, these results show that not only GRK2/3 recruitment but also kinase activity drives the recruitment of ␤arr2 to the WT and mutant D2Rs. These results support the hypothesis that direct GRK2 recruitment to the D2R independent of G protein activation drives its engagement of the ␤arr pathway.
To investigate the contribution of the other subfamily of GRKs (GRK4/5/6), we also performed the same assay with overexpression of GRK6 (Fig. 3, G-I). We did observe slight, statistically significant effects at each D2R mutant; however, the results are in contrast to the much larger effect we observed previously with GRK2 overexpression (28), where we observed a full log increase in ␤arr2 recruitment potency at the [WT] D2R that was reversed by Cmpd101. Therefore, we conclude that GRK2 is the key GRK necessary for ␤arr2 recruitment by the D2R.

Comparison of agonist-induced D2R phosphorylation with PTX or Cmpd101 pretreatment
We next used 32 P metabolic labeling to compare the effect of PTX or Cpmd101 pretreatment on DA-induced phosphorylation of the D2Rs. As shown in Fig. 4A, we observed no signal in our assay under the mock-transfected condition, indicating that our assay was measuring 32 P incorporated into the transfected D2Rs. In Fig. 4B, we observed a statistically significant effect of DA stimulation alone versus the unstimulated condition at the [WT] D2R and [Gprot] D2R but, surprisingly, not at the [␤arr] D2R. However, this does not necessarily mean that the [␤arr] D2R does not undergo DA-induced phosphorylation. As reported previously (38), the effect of DA alone was rather modest (50% above unstimulated), and we do not know which residues are phosphorylated. Nevertheless, these results do demonstrate that, when pretreated with Cmpd101 or PTX, the effect of DA no longer differed from the unstimulated condition for any D2R (Fig. 4B), indicating that both direct GRK2 kinase inhibition (Cmpd101) or pretreatment with PTX can reduce D2R phosphorylation.   D2R We next directly examined the role of G protein-dependent mechanisms in the recruitment of GRK2 to the D2Rs. By using a BRET assay for GRK2 recruitment (as schematically shown in Fig. 2A), we determined that PTX pretreatment can significantly reduce the recruitment of GRK2 to the [WT] D2R (Fig.  5A). As expected based on our previous data, there was no effect of PTX on GRK2 recruitment at the [␤arr] D2R (Fig. 5B). We also did not observe an effect of PTX at the [Gprot] D2R (Fig. 5C). These results are similar to the previous results assessing the effect of PTX on ␤arr2 recruitment.
As a control, we also assess the effect of Cmpd101 (10 M) on GRK2 recruitment to the D2Rs (Fig. 5, D-F). As expected for a kinase inhibitor, Cmpd101 had no effect on GRK2 recruitment at any D2R.
We then tested whether a mutant GRK2 that lacks the C-terminal G␤␥ binding domain could still be recruited to the D2Rs. We engineered a truncated form of the GRK2-eYFP construct that eliminates the last 28 amino acids from GRK2 before its in-frame fusion with eYFP (Fig. 5G). The Val-661-truncated GRK2 mutant has been shown previously to disrupt G␤␥-mediated translocation of GRK2 without affecting kinase activity (42).
We next assessed the interaction between the Val-661truncated GRK2 BRET construct with the D2Rs. As predicted based on our previous results, we found that this GRK2 mutant was able to be dose-dependently recruited to the [WT] D2R and the [␤arr] D2R (Fig. 5H). The largest degree of recruitment occurred at the [␤arr] D2R, which supports the hypothesis that the [␤arr] D2R can recruit GRK2 completely independently of G protein-dependent mechanisms. The [Gprot] D2R caused relatively little mutant GRK2 recruitment, which highlights the fact that GRK2 recruitment occurs in both a G protein-dependent and -independent manner. Taken together, these results clearly demonstrate that the D2R can directly recruit GRK2 independent of G protein mechanisms.

Stimulation of the mutant D2Rs with UNC9994 demonstrates an additive effect between [␤arr] D2R and UNC9994
Because the agonist activity of UNC9994 at the ␤arr pathway appears to be dependent on GRK2 (28), we next tested its effect at the biased D2R mutants. We hypothesized that ␤arr2 recruitment induced by UNC9994 at the [␤arr] D2R would be greater than at the [WT] D2R. Indeed, we find that, in HEK293T cells with endogenous GRK2 levels, UNC9994 has a significantly increased ability to recruit ␤arr2 at the [␤arr] D2R compared with either [WT] D2R or [Gprot] D2R (Fig. 6A). This could be predicted based on both the GRK2 dependence of UNC9994 and the increased G protein-independent recruitment of GRK2 to the [␤arr] D2R (28,32). When GRK2 was overexpressed (Fig. 6B), UNC9994 could cause increased ␤arr2 recruitment to the [WT] D2R, as observed previously (28). This effect was also apparent at the [␤arr] D2R and, to a smaller degree, at the [Gprot] D2R. As a test for the kinase specificity, we also overexpressed GRK6 (Fig. 6C) and found essentially no effect at

Direct GRK2 recruitment by the D2R
the [WT] D2R, although we did observe statistically significant effects at the mutant D2Rs compared with the WT, but not to the level of GRK2 overexpression.
We next compared the ability of UNC9994 to induce GRK2 recruitment to the WT and mutant D2Rs. We found that GRK2 recruitment was significantly enhanced at the [␤arr] D2R (Fig.  6D). Again, this illustrates a case where a ␤arr-biased compound is engaging the ␤arr pathway through GRK2 in a G protein-independent manner. Taken together, these data show that UNC9994-induced ␤arr2 recruitment depends on the inherent ability of the particular D2R to recruit GRK2.

Discussion
The GPCR desensitization program enables cells to appropriately respond to high ligand concentrations by down-regulating receptors. This linear pathway begins with G protein activation and terminates after ␤arr binding. ␤arr binding prevents G protein overactivation while also transducing G protein-independent signaling. GRK recruitment and activity link G protein activation to ␤arr recruitment. Therefore, designing functionally selective ligands that solely activate either the G protein or␤arrsignalingarmsischallengingbecauseoftheinterdependence of these signaling events. Identifying receptor and

Direct GRK2 recruitment by the D2R
pharmacological mechanisms that disentangle these pathways should lead to better drugs targeting GPCRs.
Here we investigated how the D2R can achieve ␤arr recruitment via direct engagement and activation of GRK2 independent of activated G proteins. We found that the [WT] D2R and biased D2R mutants (to varying degrees) can recruit GRK2 directly with a rather modest involvement of G␤␥-mediated GRK2 translocation. In fact, the [␤arr] D2R robustly recruits GRK2 without activating G proteins (Fig. 1, B and C), and therefore G protein inhibition has no effect on its recruitment of ␤arr2 or GRK2 (Figs. 2C and 5B). We have also confirmed a key role for GRK2 kinase activity at recruiting ␤arr2 to the D2R, as GRK2 inhibition pharmacologically or genetically reduces ␤arr2 recruitment at the D2R (Fig. 3). These results demonstrate that direct recruitment of GRK2 can play a key role during ␤arr-biased activation at the D2R. A caveat to these results is that they were obtained using overexpressed D2Rs, which may overemphasize the importance of G protein-independent GRK2 recruitment at the D2R. Therefore, we characterized the response of the biased D2R mutants to the ␤arr-biased D2R ligand UNC9994. We have observed previously that UNC9994's degree of activity at D2R/ ␤arr2 is strikingly dependent on the expression levels of GRK2 both in vitro and in vivo (28). In the context of this work, the increased ability of the [␤arr] D2R to directly recruit GRK2 enables UNC9994 to cause recruitment of ␤arr2 to the [␤arr] D2R, even at endogenous GRK2 expression levels (Fig.  6A). This demonstrates that direct GRK2 recruitment is relevant to achieving ␤arr-biased signaling and suggests that this mechanism could be further exploited by forthcoming D2R ligands.
In conclusion, this study enhances our understanding of how a ␤arr-biased D2R mutant achieves its signaling profile through direct, G protein-independent recruitment and activation of GRK2. We also demonstrate that the D2R can recruit GRK2 with an unexpectedly large G protein-independent component. This work also provides proof of concept that direct recruitment of GRK2 to the D2R could be a means of tailoring the profile of a functionally selective D2R ligand. Often, the key contribution of GRKs to mediating the switch between G protein and ␤arr engagement is overlooked when screening for or assessing biased drugs. Structural work investigating the mechanism of GRK5 activation by GPCRs and investigations of GRK2 activation (23,24) indicate that GRK activation is a key feature of GPCR signal transduction. Therefore, elucidating this and other mechanisms of selective G protein or ␤arr pathway engagement will inform future efforts to design functionally selective GPCR therapeutics.

Plasmids
For the BRET experiments, previous vectors encoding the Mus musculus dopamine D2 receptor (long variant) in either the WT, Gprot, ␤arr, or D80A mutants was cloned in-frame the N1-Rluc vector (PerkinElmer Life Sciences) (26). We modified the BRET constructs by replacing Rluc with RlucII on the D2Rs. The Rluc was mutated into RlucII by introducing two point mutations (C124A and M185V) via PCR mutagenesis. Subsequently, all DNA vectors were resequenced via Sanger sequencing (Eton Bioscience Inc., San Diego, CA) to verify the correct identity of the D2R mutants and to check for random mutations.
Additionally, the sensitivity of the D2R/␤arr2 BRET experiments was further increased by cloning M. musculus ␤arr2 into the N1-mVenus vector (Addgene plasmid 54640) instead of the previously used N1-eYFP vector (Clontech, Mountain View, CA). This was done by PCR amplification of ␤arr2 with the following primers to add a 5Ј NheI site (5Ј-GCTAGCA-

Direct GRK2 recruitment by the D2R
TGGGAGAAAAACCCGGGACCAGGG) and a 3Ј ApaI site (5Ј-GGGCCCGGCAGAACTGGTCATCACAGTCATC) and cutting this PCR product into the linearized N1-mVenus vector. This resulted in about a 2-fold increase in the maximal BRET signal versus the Rluc/eYFP construct pair (data not shown).
For the V661 truncation GRK2-eYFP vector, Bos taurus WT GRK2 was PCR-amplified with the following primers to add 5Ј HindIII and 3Ј BamHI sites immediately following Val-661 with the addition of GC to maintain the correct reading frame: 5Ј-CGAGCTCAAGCTTCCAATTCGGCGCCATGGCGGACC-TGGAGGCGGTGC and 3Ј-GGATCCCGGTTCTTCATCT-TGGGCACCCG. This Val-661-truncated GRK2 PCR product was cloned into the same N1-eYFP vector that was used for the WT GRK2. The subsequent clones were verified by Sanger sequencing.
For the TGF␣ shedding assay, the plasmids were the same as those described in Ref. 29, except for the D2Rs, which were the same as in Ref. 26. For the D2R phosphorylation assay, the plasmids were the same as in Ref. 26.

Direct GRK2 recruitment by the D2R Cells
HEK293T cells obtained from the ATCC were cultured and transfected via the standard calcium phosphate method and were used in all BRET assays. For the TGF␣ shedding assay, we used HEK293 cells that lack G q/11 , which were previously modified via CRISPR/Cas9 and characterized (44) to reduce nonspecific shedding activity.

Chemicals and drugs
Dopamine hydrochloride, ascorbic acid, 4-methylumbelliferyl phosphate (4-MUP) disodium salt, and DMSO were purchased from Sigma-Aldrich (St. Louis, MO). Cmpd101 was purchased from HelloBio (Princeton, NJ). Coelenterazine h and pertussis toxin were purchased from Cayman Chemical (Ann Arbor, MI). Hanks' balanced salt solution (HBSS) with Ca 2ϩ and Mg 2ϩ was purchased from Thermo Fisher Scientific (Waltham, MA). UNC9994 was synthesized as described previously (27). Phosphorus-32 radionuclide orthophosphoric acid in water (specific activity, 285.6 Ci/mg; concentration, 10 mCi/ ml) was purchased from PerkinElmer Life Sciences (Waltham, MA). Unless otherwise stated, all other incidental chemicals were purchased from Sigma-Aldrich. Cmpd101 was initially dissolved in DMSO at 10 mM and subsequently diluted in HBSS. UNC9994 was initially dissolved in DMSO at 10 mM and subsequently diluted in drug dilution buffer (HBSS supplemented with 20 mM HEPES, 0.3% BSA, 0.03% ascorbic acid) to maintain the solubility of the compound. Dopamine was dissolved directly in HBSS supplemented with 0.03% ascorbic acid to prevent oxidation, except where compared directly with UNC9994, where the drug dilution buffer was used instead.

BRET assays
BRET assays were performed as described previously using predetermined ratios of RlucII and mVenus-tagged proteins (31). Briefly, HEK293T cells were seeded at 70% confluence in a 6-well plate on the previous day and transfected via the calcium phosphate method using 0.4 g of D2R-RlucII and either 2.5 g of ␤arr2-mVenus or 1.0 g of GRK2-eYFP. The following day, cells were plated onto poly-D-lysine-coated, clear-bottom, white-walled 96-well plates using clear minimum Eagle's medium supplemented with 10 mM HEPES, 1ϫ GlutaMax, 2% fetal bovine serum, and 1ϫ Anti-Anti (BRET medium). When cells were treated with PTX, they were incubated with 200 ng/ml overnight before the day of the experiment. When cells were treated with Cmpd101, cells were incubated with 10 M for 1 h prior to running the assay. On the day of the assay, the medium was removed, a white vinyl sticker was placed on the bottom of the plate, and HBSS was added to bring the final volume to 100 l per well. A 10ϫ concentration of coelenterazine h (10 l; final concentration, ϳ4.7 M) along with a 10ϫ dose response of ligand (10 l) was added, and the plates were read on a Berthold Mithras LB 940 plate reader (Bad Wildbad, Germany). These readings were averaged per plate, and the plate averages were combined in the graphs presented with the basal BRET ratio (unstimulated) subtracted to give the net

Direct GRK2 recruitment by the D2R
BRET ratio. There was no significant difference between the D2R mutants at baseline in any of the assays.

TGF␣ shedding assay
Performed essentially as described previously (29) with a difference in the alkaline phosphatase substrate and the use of G q/11 CRISPR knockout cells (44). We used the fluorescent substrate 4-MUP, instead of p-nitrophenyl phosphate. This enabled us to measure the alkaline phosphatase using a ClarioStar plate reader (BMG Labtech, Ortenberg, Germany) equipped with a fluorescent reader, with excitation set at 360 nm (Ϯ 10 nm) and emission at 450 nm (Ϯ 15 nm). Data were collected over the course of 30 min. The shedding activity was calculated by dividing the amount of phosphatase activity present in the conditioned medium by the amount present on the cells plus the conditioned medium. These values were then vehicle-subtracted and normalized (percent) to the amount of alkaline phosphatase activity in the conditioned medium induced by 100 nM 12-O-tetradecanoylphorbol-13-acetate for each transfection condition.

D2R phosphorylation experiments
These experiments were performed essentially as described previously (38,45) with slight modifications. 3ϫHA epitopetagged D2R receptor constructs were transfected into HEK293T cells. The next day, cells were replated at high density onto poly-D-lysine-coated 6-well plates in low-serum BRET medium. When cells were treated with PTX, they were incubated overnight with 200 ng/ml. The following day, cells were metabolically labeled with 100 Ci of 32 P i in phosphate-free Dulbecco's modified Eagle's medium containing 20 mM HEPES and antibiotics/antimitotics at 37°C for 60 min. Cmpd101 (10 M) or vehicle (diluted DMSO) was included during the metabolic labeling. Samples were then stimulated with 10 M DA or vehicle for 10 min at 37°C. Then cells were rinsed in ice-cold PBS and lysed with RIPAϩ buffer (150 mM NaCl, 50 mM Tris, 5 mM EDTA, 1% (v/v) NP-40, 0.5% (w/v) deoxycholate (sodium salt), 0.1% (w/v) SDS, 50 mM NaF, and 10 mM sodium pyrophosphate with supplemented with fresh EDTA-free protease inhibitor mixture). Samples were then solubilized by rotating for 45 min at 4°C. Samples were then spun down to remove insoluble debris and transferred to a new tube containing 15 l of anti-HA magnetic beads (Pierce anti-HA magnetic beads, catalog no. 88836) and incubated with rotation for 1.5 h at 4°C. Beads were then washed twice with ice-cold RIPAϩ buffer and then once with TBS-T. Then samples were eluted at 65°C for 10 min with Laemmli sample buffer containing 100 mM DTT.
Samples were then loaded to a 4 -12% BisTris gel for SDS-PAGE analysis (Thermo Fisher Scientific, catalog no. NP0323BOX). After mass separation on the gel, samples were transferred to polyvinylidene difluoride membranes (Thermo Scientific, catalog no. 88585), and then the membranes were dried and subjected to autoradiography. Membranes were placed in cassettes containing intensifying screens (Kodak Bio-Max MS) and film (Kodak BioMax MR film). The cassettes were placed at Ϫ80°C for 48 -72 h before film development. The films were scanned using a film scanner, and images were quantified using ImageJ (National Institutes of Health). Following this, the same membranes were subjected to a Western blot procedure using a rabbit anti-HA tag primary antibody (Cell Signaling Technology, catalog no. 3724) and Li-Cor goat antirabbit 800 secondary (Li-Cor, catalog no. 926-32211). Membranes were then scanned on the Li-Cor near-IR scanner, and images were quantified using ImageJ. Values were determined by normalizing the autoradiography signal to the anti-HA Western blot for the same region and then normalizing to the unstimulated condition for each D2R mutant.

Curve-fitting and statistical analysis
Dose-response curves were fit using GraphPad Prism's log (agonist) versus response (three parameters) nonlinear fit function to calculate logEC 50 and maximum efficacy values. Statistical analysis was performed in Prism version 7.0 (GraphPad Software Inc., La Jolla, CA) as indicated in the text. Experiments testing the effect of a manipulation at each receptor were considered as one statistical unit for the purpose of ANOVA, which was performed before Bonferroni-corrected t tests. Each BRET or TGF shedding assay was performed in duplicate with at least four independent replicates.